sdk/emulator/qemu.git
14 years agoeepro100: fix PCI interrupt pin configuration regression
Stefan Weil [Tue, 6 Apr 2010 11:44:09 +0000 (13:44 +0200)]
eepro100: fix PCI interrupt pin configuration regression

Commit 15e89f5916c9e82347cbd1fd416db3e348bab426
removed this setting, but it is still needed.

Without this patch, e100 device drivers using
interrupts don't work with qemu.

See other nic emulations which also set the
PCI interrupt pin.

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agoeepro100: fix mapping of flash memory
Stefan Weil [Tue, 6 Apr 2010 11:44:08 +0000 (13:44 +0200)]
eepro100: fix mapping of flash memory

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agoeepro100: Set power management capability using pci_reserve_capability
Stefan Weil [Tue, 6 Apr 2010 11:44:07 +0000 (13:44 +0200)]
eepro100: Set power management capability using pci_reserve_capability

pci_add_capability automatically updates PCI status and
PCI capability pointer, so use it. Use pci_reserve_capability
to make the new capability appear at the correct offset.

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agoeepro100: Set configuration bit for standard TCB
Stefan Weil [Tue, 6 Apr 2010 11:44:05 +0000 (13:44 +0200)]
eepro100: Set configuration bit for standard TCB

For some devices, this bit is always set.
For the others, it is set by default.

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agoeepro100: Add new device variant i82801
Stefan Weil [Tue, 6 Apr 2010 11:44:04 +0000 (13:44 +0200)]
eepro100: Add new device variant i82801

This ethernet device is used in Toshiba Tecra 8200 notebooks.

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agoeepro100: Simplified device instantiation
Stefan Weil [Tue, 6 Apr 2010 11:44:03 +0000 (13:44 +0200)]
eepro100: Simplified device instantiation

By using a private device info structure
(as suggested by Gerd Hoffmann), handling of the
different device variants becomes much easier.

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agoeepro100: Simplify status handling
Stefan Weil [Tue, 6 Apr 2010 11:44:02 +0000 (13:44 +0200)]
eepro100: Simplify status handling

Includes a minor STATUS_NOT_OK -> 0 tweak.

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agoeepro100: Don't allow writing SCBStatus
Stefan Weil [Tue, 6 Apr 2010 11:44:01 +0000 (13:44 +0200)]
eepro100: Don't allow writing SCBStatus

SCBStatus is readonly, but most drivers which were derived
from the old Linux eepro100.c do a word write to this address
when they want to acknowledge interrupts.

So we have to mask these writes here.

The patch also removes old unused code for status read / write.

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agotcg/ppc: Fix typo
malc [Mon, 5 Apr 2010 23:10:03 +0000 (03:10 +0400)]
tcg/ppc: Fix typo

Signed-off-by: malc <av1474@comtv.ru>
14 years agotcg/ppc: Implment bswap16/32
malc [Mon, 5 Apr 2010 22:53:11 +0000 (02:53 +0400)]
tcg/ppc: Implment bswap16/32

Signed-off-by: malc <av1474@comtv.ru>
14 years agoOHCI qdev conversion
Paul Brook [Mon, 5 Apr 2010 18:57:40 +0000 (19:57 +0100)]
OHCI qdev conversion

Convert remaining OHCI devices to QDEV interface.

Signed-off-by: Paul Brook <paul@codesourcery.com>
14 years agoFix arm-linux-user
Paul Brook [Mon, 5 Apr 2010 18:56:34 +0000 (19:56 +0100)]
Fix arm-linux-user

Only include hw/loader.h from target-arm/helper.c when building for
system emulation.

Signed-off-by: Paul Brook <paul@codesourcery.com>
14 years agoARMv7-M reset fixes
Paul Brook [Mon, 5 Apr 2010 18:34:51 +0000 (19:34 +0100)]
ARMv7-M reset fixes

Move ARMv7-M PC/SP initialization to the CPU reset routine.  Add a board
reset routine to call this.  Also load values directly from ROM as
images have not been copied yet.

Avoid clearing the NVIC pointer on cpu reset.

Signed-off-by: Paul Brook <paul@codesourcery.com>
14 years agotcg/mips: use seb/seh instructions on MIPS32R2
Aurelien Jarno [Mon, 5 Apr 2010 11:19:56 +0000 (13:19 +0200)]
tcg/mips: use seb/seh instructions on MIPS32R2

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotcg/mips: fix 64-bit linux-user on big endian MIPS
Aurelien Jarno [Mon, 5 Apr 2010 12:49:27 +0000 (14:49 +0200)]
tcg/mips: fix 64-bit linux-user on big endian MIPS

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotcg/ppc: Implement eqv, nand and nor
malc [Mon, 5 Apr 2010 12:08:47 +0000 (16:08 +0400)]
tcg/ppc: Implement eqv, nand and nor

Signed-off-by: malc <av1474@comtv.ru>
14 years agoSplit TLB addend and target_phys_addr_t
Paul Brook [Sun, 4 Apr 2010 23:28:53 +0000 (00:28 +0100)]
Split TLB addend and target_phys_addr_t

Historically the qemu tlb "addend" field was used for both RAM and IO accesses,
so needed to be able to hold both host addresses (unsigned long) and guest
physical addresses (target_phys_addr_t).  However since the introduction of
the iotlb field it has only been used for RAM accesses.

This means we can change the type of addend to unsigned long, and remove
associated hacks in the big-endian TCG backends.

We can also remove the host dependence from target_phys_addr_t.

Signed-off-by: Paul Brook <paul@codesourcery.com>
14 years agoUHCI spurious interrut fix
Paul Brook [Sun, 4 Apr 2010 20:48:31 +0000 (21:48 +0100)]
UHCI spurious interrut fix

Only raise an interrupt if the TD has actually completed.

Signed-off-by: Paul Brook <paul@codesourcery.com>
14 years agoRevert "Compile usb-ohci only once"
Paul Brook [Sun, 4 Apr 2010 20:18:26 +0000 (21:18 +0100)]
Revert "Compile usb-ohci only once"

This reverts commit f1698408f1dcb7548a21828a0b1e2b530fae3af3.

PCI is always little-endian. Having a user-visible "be" property is just
plain wrong.

14 years agotcg/ppc: Fix not_i32
malc [Sun, 4 Apr 2010 16:36:29 +0000 (20:36 +0400)]
tcg/ppc: Fix not_i32

Thanks to Alexander Graf for bug report and a good reproducible test
case.

Signed-off-by: malc <av1474@comtv.ru>
14 years agoMake cpu_get_real_ticks use mfspr
Alexander Graf [Sat, 3 Apr 2010 09:37:26 +0000 (11:37 +0200)]
Make cpu_get_real_ticks use mfspr

PowerPC CPUs have had two ways to read the time base for quite some time now.
They provide it using the mfspr instruction or - if a special bit is set in
that opcode - using mftb. For timekeeping we're currently using mftb.

While trying to get Qemu up and running on an e500v2 system, I stumbled over
the CPU not supporting mftbu. It just throws an illegal instruction trap.

So let's read the SPR values instead. All PPC CPUs should support them anyways.

I tested this patch on an e500v2 system where it makes qemu work and on my 970MP
system with 32-bit user space where everything still works with this patch
applied.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: malc <av1474@comtv.ru>
14 years agosparc32: add IOMMU chipset doc links
Blue Swirl [Sat, 3 Apr 2010 07:40:47 +0000 (07:40 +0000)]
sparc32: add IOMMU chipset doc links

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agosparc32: rename iommu.c to sun4m_iommu.c to make room for other IOMMUs
Blue Swirl [Sat, 3 Apr 2010 07:35:50 +0000 (07:35 +0000)]
sparc32: rename iommu.c to sun4m_iommu.c to make room for other IOMMUs

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agosparc32: improve timer implementation
Blue Swirl [Sat, 3 Apr 2010 06:17:35 +0000 (06:17 +0000)]
sparc32: improve timer implementation

Timer with zero period (free-run) will never match.

Timer counting starts with tick value of 0x200, not from 0,
so the period must calculated from one tick less than the limit.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoFix build on mingw32
Anthony Liguori [Fri, 2 Apr 2010 15:26:30 +0000 (10:26 -0500)]
Fix build on mingw32

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agohw/r2d: add a USB keyboard
Aurelien Jarno [Fri, 2 Apr 2010 10:16:04 +0000 (12:16 +0200)]
hw/r2d: add a USB keyboard

The R2D board does not have a PS/2 port, and only support a keyboard on
the USB bus.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agousb-bus: fix no params
TeLeMan [Tue, 30 Mar 2010 01:33:24 +0000 (09:33 +0800)]
usb-bus: fix no params

After commit 702f3e0fb52c124c07f215426eeadb70a716643f, the params is
nerver NULL. It should check *params instead of params to determine
whether the params is empty.

Signed-off-by: TeLeMan <geleman@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoUpdate PowerPC OpenBIOS image to r721
Aurelien Jarno [Fri, 2 Apr 2010 08:07:44 +0000 (10:07 +0200)]
Update PowerPC OpenBIOS image to r721

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agowin32: Fix compiler errors from u_int64_t
Stefan Weil [Thu, 1 Apr 2010 21:59:51 +0000 (16:59 -0500)]
win32: Fix compiler errors from u_int64_t

u_int64_t raises compiler error messages:

  CC    libhw32/virtio.o
/qemu/ar7/hw/virtio.c: In function ‘virtio_queue_get_avail_size’:
/qemu/ar7/hw/virtio.c:776: error: ‘u_int64_t’ undeclared (first use in this function)
/qemu/ar7/hw/virtio.c:776: error: (Each undeclared identifier is reported only once
/qemu/ar7/hw/virtio.c:776: error: for each function it appears in.)

Replacing u_int64_t by uint64_t helps.

Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agotcg/TODO: remove setcond
Aurelien Jarno [Thu, 1 Apr 2010 20:00:41 +0000 (22:00 +0200)]
tcg/TODO: remove setcond

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoFix compilation with missing inotify_init1
Stefan Weil [Sun, 28 Mar 2010 09:44:41 +0000 (11:44 +0200)]
Fix compilation with missing inotify_init1

Commit c05c7a7306a23a4b01d1606172b142c45caffc92
breaks cross compilation for mips (and other
compilations without CONFIG_INOTIFY1):

make[1]: Entering directory `/qemu/bin/mips'
  CC    i386-linux-user/syscall.o
cc1: warnings being treated as errors
/qemu/linux-user/syscall.c: In function ‘do_syscall’:
/qemu/linux-user/syscall.c:7067: error: implicit declaration of function ‘sys_inotify_init1’

Cc: Riku Voipio <riku.voipio@nokia.com>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoS390: Tell user why VM creation failed
Alexander Graf [Thu, 1 Apr 2010 16:42:37 +0000 (18:42 +0200)]
S390: Tell user why VM creation failed

The KVM kernel module on S390 refuses to create a VM when the switch_amode
kernel parameter is not used.

Since that is not exactly obvious, let's give the user a nice warning.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoS390: Add stub for cpu_get_phys_page_debug
Alexander Graf [Thu, 1 Apr 2010 16:42:36 +0000 (18:42 +0200)]
S390: Add stub for cpu_get_phys_page_debug

We don't implement any virtual memory in the S390 target so far, so let's
add a stub for this now mandatory function.

Fixes building of S390 target.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotcg: initial ia64 support
Aurelien Jarno [Mon, 29 Mar 2010 00:12:51 +0000 (02:12 +0200)]
tcg: initial ia64 support

A few words about design choices:
* On IA64, instructions should be grouped by bundle, and dependencies
  between instructions declared. A first version of this code tried to
  schedule instructions automatically, but was very complex and too
  invasive for the current common TCG code (ops not ending at
  instruction boundaries, code retranslation breaking already generated
  code, etc.)  It was also not very efficient, as dependencies between
  TCG ops is not available.
  Instead the option taken by the current implementation does not try
  to fill the bundle by scheduling instructions, but by providing ops
  not available as an ia64 instruction, and by offering 22-bit constant
  loading for most of the instructions. With both options the bundle are
  filled at approximately the same level.

* Up to 128 registers can be affected to a function on IA64, but TCG
  limits this number to 64, which is actually more than enough. The
  register affectation is the following:
  - r0: used to map a constant argument with value 0
  - r1: global pointer
  - r2, r3: internal use
  - r4 to r6: not used to avoid saving them
  - r7: env structure
  - r8 to r11: free for TCG (call clobbered)
  - r12: stack pointer
  - r13: thread pointer
  - r14 to r31: free for TCG (call clobbered)
  - r32: reserved (return address)
  - r33: reserved (PFS)
  - r33 to r63: free for TCG

* The IA64 architecture has only 64-bit registers and no 32-bit
  instructions (the only exception being cmp4). Therefore 64-bit
  registers and instructions are used for 32-bit ops. The adopted
  strategy is the same as the ABI, that is the higher 32 bits are
  undefined. Most ops (and, or, add, shl, etc.) can directly use
  the 64-bit registers, while some others have to sign-extend (sar,
  div, etc.) or zero-extend (shr, divu, etc.) the register first.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotcg: align static_code_gen_buffer to CODE_GEN_ALIGN
Aurelien Jarno [Mon, 29 Mar 2010 00:12:51 +0000 (02:12 +0200)]
tcg: align static_code_gen_buffer to CODE_GEN_ALIGN

On ia64, the default memory alignement is not enough for a code
alignement. To fix that, force static_code_gen_buffer alignment
to CODE_GEN_ALIGN.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoia64 disas support
Aurelien Jarno [Mon, 29 Mar 2010 00:12:51 +0000 (02:12 +0200)]
ia64 disas support

Taken from binutils SVN, using last GPLv2 version.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agolinux-user: fix page_unprotect when host page size > target page size
Aurelien Jarno [Mon, 29 Mar 2010 00:12:51 +0000 (02:12 +0200)]
linux-user: fix page_unprotect when host page size > target page size

When the host page size is bigger that the target one, unprotecting a
page should:
- mark all the target pages corresponding to the host page as writable
- invalidate all tb corresponding to the host page (and not the target
  page)

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agolinux-user/ia64: workaround ia64 strangenesses
Aurelien Jarno [Mon, 29 Mar 2010 00:12:51 +0000 (02:12 +0200)]
linux-user/ia64: workaround ia64 strangenesses

ia64 has some strangenesses that need to be workaround:
- it has a __clone2() syscall instead of the using clone() one, with
  different arguments, and which is not declared in the usual headers.
- ucontext.uc_sigmask is declared with type long int, while it is
  actually of type sigset_t.
- uc_mcontext, uc_sigmask, uc_stack, uc_link are declared using #define,
  which clashes with the target_ucontext fields. Change their names to
  tuc_*, as already done for some target architectures.

14 years agovirtio-net: vhost net support
Michael S. Tsirkin [Wed, 17 Mar 2010 11:08:42 +0000 (13:08 +0200)]
virtio-net: vhost net support

This connects virtio-net to vhost net backend.
The code is structured in a way analogous to what we have with vnet
header capability in tap.

We start/stop backend on driver start/stop as
well as on save and vm start (for migration).

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agotap: add API to retrieve vhost net header
Michael S. Tsirkin [Wed, 17 Mar 2010 11:08:38 +0000 (13:08 +0200)]
tap: add API to retrieve vhost net header

will be used by virtio-net for vhost net support

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agotap: add vhost/vhostfd options
Michael S. Tsirkin [Wed, 17 Mar 2010 11:08:24 +0000 (13:08 +0200)]
tap: add vhost/vhostfd options

This adds vhost binary option to tap, to enable vhost net accelerator.
Default is off for now, we'll be able to make default on long term
when we know it's stable.

vhostfd option can be used by management, to pass in the fd. Assigning
vhostfd implies vhost=on.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovhost: vhost net support
Michael S. Tsirkin [Wed, 17 Mar 2010 11:08:17 +0000 (13:08 +0200)]
vhost: vhost net support

This adds vhost net device support in qemu. Will be tied to tap device
and virtio by following patches.  Raw backend is currently missing,
will be worked on/submitted separately.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovirtio-pci: fill in notifier support
Michael S. Tsirkin [Wed, 17 Mar 2010 11:08:13 +0000 (13:08 +0200)]
virtio-pci: fill in notifier support

Support host/guest notifiers in virtio-pci.
The last one only with kvm, that's okay
because vhost relies on kvm anyway.

Note on kvm usage: kvm ioeventfd API
is implemented on non-kvm systems as well,
this is the reason we don't need if (kvm_enabled())
around it.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovirtio: move typedef to qemu-common
Michael S. Tsirkin [Wed, 17 Mar 2010 11:08:10 +0000 (13:08 +0200)]
virtio: move typedef to qemu-common

make it possible to use type without header include,
simplifying header dependencies.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovirtio: add set_status callback
Michael S. Tsirkin [Wed, 17 Mar 2010 11:08:05 +0000 (13:08 +0200)]
virtio: add set_status callback

vhost net backend needs to be notified when
frontend status changes. Add a callback,
similar to set_features.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovirtio: notifier support + APIs for queue fields
Michael S. Tsirkin [Wed, 17 Mar 2010 11:08:02 +0000 (13:08 +0200)]
virtio: notifier support + APIs for queue fields

vhost needs physical addresses for ring and other queue fields,
so add APIs for these. In particular, add binding API to set
host/guest notifiers.  Will be used by vhost.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agonotifier: event notifier implementation
Michael S. Tsirkin [Wed, 17 Mar 2010 11:07:58 +0000 (13:07 +0200)]
notifier: event notifier implementation

event notifiers are slightly generalized eventfd descriptors. Current
implementation depends on eventfd because vhost is the only user, and
vhost depends on eventfd anyway, but a stub is provided for non-eventfd
case.

We'll be able to further generalize this when another user comes along
and we see how to best do this.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agokvm: add API to set ioeventfd
Michael S. Tsirkin [Wed, 17 Mar 2010 11:07:54 +0000 (13:07 +0200)]
kvm: add API to set ioeventfd

Comment on kvm usage: rather than require users to do if (kvm_enabled())
and/or ifdefs, this patch adds an API that, internally, is defined to
stub function on non-kvm build, and checks kvm_enabled for non-kvm
run.

While rest of qemu code still uses if (kvm_enabled()), I think this
approach is cleaner, and we should convert rest of code to it
long term.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agotap: add interface to get device fd
Michael S. Tsirkin [Wed, 17 Mar 2010 11:07:50 +0000 (13:07 +0200)]
tap: add interface to get device fd

Will be used by vhost to attach/detach to backend.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovirtio-pci: compile per-target
Anthony Liguori [Wed, 31 Mar 2010 16:52:44 +0000 (11:52 -0500)]
virtio-pci: compile per-target

With vhost, virtio-pci needs to include kvm.h and kvm.h needs to be built
per-target.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoFix -enable-kvm
Anthony Liguori [Thu, 1 Apr 2010 13:33:06 +0000 (08:33 -0500)]
Fix -enable-kvm

Make vl.o compiled per target and fix a thinko in hw/acpi.c.  It's not trivial
to make kvm.h consumable by compiled-once files.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoRestore terminal monitor attributes - addition
Shahar Havivi [Sun, 14 Mar 2010 20:41:15 +0000 (22:41 +0200)]
Restore terminal monitor attributes - addition

Patch 2d753894c7553d6a05e8fdbed5f4704398919a35 was missing this check,
when running monitor as /dev/tty and other serial device, i.e:
  qemu -monitor /dev/tty -serial /dev/pts/1

Without this patch any serial device will override the monitor stored
attributes. (monitor is called in main() before any serial device).

Signed-off-by: Shahar Havivi <shaharh@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoDon't check for bus master for old guests
Alexander Graf [Tue, 16 Mar 2010 18:18:07 +0000 (19:18 +0100)]
Don't check for bus master for old guests

Older Linux guests don't activate the bus master enable bit. So for those we
can just try to be clever and track if they set the DEVICE_OK bit even though
bus mastering is still disabled.

Under that condition we can disable the windows safety check. With that logic
in place both guests should work just fine. Without PCI hotplug breaks
virtio-net in Linux < 2.6.34 guests.

Signed-off-by: Alexander Graf <agraf@suse.de>
CC: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agousb: class specific interface requests
Max Reitz [Sun, 14 Mar 2010 11:19:03 +0000 (12:19 +0100)]
usb: class specific interface requests

Mass Storage Reset and Get Max LUN are class specific requests, but
they were not marked as such in hw/usb-msd.c, moved therefore
ClassInterfaceRequest and ClassInterfaceOutRequest from hw/usb-net.c
to hw/usb.h.
Furthermore there was a problem in hw/usb-ohci.c when using DEBUG
concerning systems where size_t is a 32 bit integer (printf resulted
in a segmentation fault).

Signed-off-by: Max Reitz <max@tyndur.org>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoAdd missing #include needed for madvise() on OpenBSD
Blue Swirl [Tue, 30 Mar 2010 19:27:34 +0000 (19:27 +0000)]
Add missing #include needed for madvise() on OpenBSD

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoFix i386-bsd-user build
Blue Swirl [Tue, 30 Mar 2010 18:24:49 +0000 (18:24 +0000)]
Fix i386-bsd-user build

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoGet bsd-user host page protection code working on FreeBSD hosts
Juergen Lock [Thu, 25 Mar 2010 21:32:16 +0000 (22:32 +0100)]
Get bsd-user host page protection code working on FreeBSD hosts

Use kinfo_getvmmap(3) on FeeBSD >= 7.x and /compat/linux/proc on older
FreeBSD.  (kinfo_getvmmap is preferred since /compat/linux/proc is
usually only mounted on hosts also using the Linuxolator.)

This patch is a bit hacky because the includes needed for kinfo_getvmmap
conflict with other definitions in exec.c by default so I had to `trick
around' a little, but I built the result in FreeBSD 6.4-stable and
7.2-stable tbs and on 8-stable on the host so the hacks at least
should be stable.  (If this is a problem maybe we could also move the
kinfo_getvmmap invocations into a seperate source file but that would
be more work...)

Signed-off-by: Juergen Lock <nox@jelal.kn-bremen.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoFix bsd-user qemu_vmalloc() host page protection code
Juergen Lock [Thu, 25 Mar 2010 21:11:17 +0000 (22:11 +0100)]
Fix bsd-user qemu_vmalloc() host page protection code

Just do the same as linux-user does.

Signed-off-by: Juergen Lock <nox@jelal.kn-bremen.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoFix some compilation warnings on FreeBSD hosts
Juergen Lock [Thu, 25 Mar 2010 21:34:00 +0000 (22:34 +0100)]
Fix some compilation warnings on FreeBSD hosts

Signed-off-by: Juergen Lock <nox@jelal.kn-bremen.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoUse sysctl instead of /proc to find executable path on FreeBSD
Juergen Lock [Thu, 25 Mar 2010 21:07:12 +0000 (22:07 +0100)]
Use sysctl instead of /proc to find executable path on FreeBSD

..since /proc usually isn't mounted on FreeBSD.

Signed-off-by: Juergen Lock <nox@jelal.kn-bremen.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoCompile most PPC devices only once
Blue Swirl [Tue, 30 Mar 2010 17:36:23 +0000 (17:36 +0000)]
Compile most PPC devices only once

Make byte swapping unconditional since PPC is big endian.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoCompile prep_pci only once
Blue Swirl [Mon, 29 Mar 2010 19:24:04 +0000 (19:24 +0000)]
Compile prep_pci only once

Make byte swapping unconditional since PPC is big endian.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoCompile qemu-timer only once
Blue Swirl [Mon, 29 Mar 2010 19:24:00 +0000 (19:24 +0000)]
Compile qemu-timer only once

Arrange various declarations so that also non-CPU code can access
them, adjust users.

Move CPU specific code to cpus.c.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoCompile openpic only once
Blue Swirl [Mon, 29 Mar 2010 19:23:59 +0000 (19:23 +0000)]
Compile openpic only once

Replace TARGET_PAGE_SIZE with 4096. Make byte swapping unconditional
since PPC is big endian.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoCompile ide/macio only once
Blue Swirl [Mon, 29 Mar 2010 19:23:57 +0000 (19:23 +0000)]
Compile ide/macio only once

Replace TARGET_PAGE_SIZE with 4096. Make byte swapping unconditional
since PPC is big endian.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoCompile pflash_cfi01 only once
Blue Swirl [Mon, 29 Mar 2010 19:23:56 +0000 (19:23 +0000)]
Compile pflash_cfi01 only once

Push TARGET_WORDS_BIGENDIAN dependency to board level.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoCompile pflash_cfi02 only once
Blue Swirl [Mon, 29 Mar 2010 19:23:55 +0000 (19:23 +0000)]
Compile pflash_cfi02 only once

Push TARGET_WORDS_BIGENDIAN dependency to board level.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoCompile acpi only once
Blue Swirl [Mon, 29 Mar 2010 19:23:52 +0000 (19:23 +0000)]
Compile acpi only once

Use qemu_irqs to trigger CMOS S3 and SMI events.

Avoid using kvm.h, which uses CPUState.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoRefactor target specific handling, compile vl.c only once
Blue Swirl [Mon, 29 Mar 2010 19:23:52 +0000 (19:23 +0000)]
Refactor target specific handling, compile vl.c only once

Move target specific functions and RAM handling to arch_init.c.

Add a flag to QEMUOptions structure to indicate for which
architectures the option is allowed, check the flag
in run time and remove conditional code in option handling.

Now that no target dependencies remain, compile vl.c only once
for all targets.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoRefactor CPUState handling out of vl.c
Blue Swirl [Mon, 29 Mar 2010 19:23:50 +0000 (19:23 +0000)]
Refactor CPUState handling out of vl.c

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoRefactor a few architecture dependent pieces in vl.c
Blue Swirl [Mon, 29 Mar 2010 19:23:50 +0000 (19:23 +0000)]
Refactor a few architecture dependent pieces in vl.c

These will be moved later.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoMove KVM and Xen global flags to vl.c
Blue Swirl [Mon, 29 Mar 2010 19:23:49 +0000 (19:23 +0000)]
Move KVM and Xen global flags to vl.c

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoMove cpu_exec_init_all() declaration to qemu-common.h
Blue Swirl [Mon, 29 Mar 2010 19:23:48 +0000 (19:23 +0000)]
Move cpu_exec_init_all() declaration to qemu-common.h

Let cpu_exec_init_all() be called from non-CPU code.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoAllow various header files to be included from non-CPU code
Blue Swirl [Mon, 29 Mar 2010 19:23:47 +0000 (19:23 +0000)]
Allow various header files to be included from non-CPU code

Allow balloon.h, gdbstub.h and kvm.h to be included from
non-CPU code.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoAdjust debug handling
Blue Swirl [Mon, 29 Mar 2010 19:23:46 +0000 (19:23 +0000)]
Adjust debug handling

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agotcg/mips: fix branch offset during retranslation
Aurelien Jarno [Mon, 29 Mar 2010 00:09:23 +0000 (02:09 +0200)]
tcg/mips: fix branch offset during retranslation

Branch offsets should only be overwritten during relocation, to support
partial retranslation.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoexec: remove dead code
Aurelien Jarno [Sun, 28 Mar 2010 16:47:25 +0000 (18:47 +0200)]
exec: remove dead code

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agolinux-user/ppc: use the Linux register layout
Rob Landley [Sun, 28 Mar 2010 14:51:43 +0000 (16:51 +0200)]
linux-user/ppc: use the Linux register layout

The dynamic linker converts the Linux layout to the AIX layout and is
reentrant so it won't do it a second time if it's already been
converted. In short it work just fine with either register layout.

OTOH, statically linked binaries expect a Linux layout.

Remove code converting the Linux layout to AIX layout so that all
binaries are presented the Linux Layout.

Signed-off-by: Rob Landley <rob@landley.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoqemu-io: fix aio help texts
Christoph Hellwig [Sun, 28 Mar 2010 10:19:31 +0000 (12:19 +0200)]
qemu-io: fix aio help texts

Fix a few typos in the help texts for the various aio commands.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotcg/arm: Replace qemu_ld32u (left over from previous commit)
Stefan Weil [Sun, 28 Mar 2010 09:44:29 +0000 (11:44 +0200)]
tcg/arm: Replace qemu_ld32u (left over from previous commit)

Commit 86feb1c860dc38e9c89e787c5210e8191800385e
did not change all occurrences of INDEX_op_qemu_ld32u
for tcg/arm.

Please note that I could not test this patch
(I have currently no arm system available).

Cc: Richard Henderson <rth@twiddle.net>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoFix driftfix option
Blue Swirl [Sat, 27 Mar 2010 21:33:46 +0000 (21:33 +0000)]
Fix driftfix option

Based on patch by Zachary Amsden.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoFix BSD and win32 builds
Blue Swirl [Sat, 27 Mar 2010 18:41:08 +0000 (18:41 +0000)]
Fix BSD and win32 builds

  CC    net/tap-bsd.o
/src/qemu/net/tap-bsd.c: In function `tap_open':
/src/qemu/net/tap-bsd.c:93: warning: implicit declaration of function `error_report'

  CC    sparc-softmmu/../net/tap-win32.o
cc1: warnings being treated as errors
/src/qemu/target-sparc/../net/tap-win32.c: In function 'net_init_tap':
/src/qemu/target-sparc/../net/tap-win32.c:709: warning: implicit declaration of function 'error_report'

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoRefactor numa mode setting
Blue Swirl [Sat, 27 Mar 2010 18:24:45 +0000 (18:24 +0000)]
Refactor numa mode setting

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agopflash_cfi02: fix incorrect TARGET_FMT_lx/d use
Blue Swirl [Sat, 27 Mar 2010 18:24:35 +0000 (18:24 +0000)]
pflash_cfi02: fix incorrect TARGET_FMT_lx/d use

Also use target_phys_addr_t for addresses.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoCompile vmware_vga only once
Blue Swirl [Sat, 27 Mar 2010 18:18:17 +0000 (18:18 +0000)]
Compile vmware_vga only once

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agotcg-mips: add guest base support
Aurelien Jarno [Sat, 27 Mar 2010 16:31:04 +0000 (17:31 +0100)]
tcg-mips: add guest base support

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotcg/mips: implement the not_i32 op the same way as gcc
Aurelien Jarno [Sat, 27 Mar 2010 15:50:55 +0000 (16:50 +0100)]
tcg/mips: implement the not_i32 op the same way as gcc

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotcg-mips: implement nor
Aurelien Jarno [Sat, 27 Mar 2010 15:32:55 +0000 (16:32 +0100)]
tcg-mips: implement nor

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agobe more specific in -mem-path error messages
Michael Tokarev [Sat, 27 Mar 2010 13:35:37 +0000 (16:35 +0300)]
be more specific in -mem-path error messages

Signed-Off-By: Michael Tokarev <mjt@tls.msk.ru>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoAdd a missing #include for FreeBSD hosts
Juergen Lock [Thu, 25 Mar 2010 21:35:03 +0000 (22:35 +0100)]
Add a missing #include for FreeBSD hosts

Signed-off-by: Juergen Lock <nox@jelal.kn-bremen.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoerror: Move qerror_report() from qemu-error.[ch] to qerror.[ch]
Markus Armbruster [Mon, 22 Mar 2010 09:29:05 +0000 (10:29 +0100)]
error: Move qerror_report() from qemu-error.[ch] to qerror.[ch]

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoerror: Link qemu-img, qemu-nbd, qemu-io with qemu-error.o
Markus Armbruster [Mon, 22 Mar 2010 09:29:04 +0000 (10:29 +0100)]
error: Link qemu-img, qemu-nbd, qemu-io with qemu-error.o

The location tracking interface is used by code shared with qemi-img,
qemu-nbd and qemu-io, so it needs to be available there.  Commit
827b0813 provides it in a rather hamfisted way: it adds a dummy
implementation to qemu-tool.c.

It's cleaner to provide the real thing, and put a few more dummy
monitor functions into qemu-tool.c.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoerror: Make use of error_set_progname() optional
Markus Armbruster [Mon, 22 Mar 2010 09:29:03 +0000 (10:29 +0100)]
error: Make use of error_set_progname() optional

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoerror: Trim includes after "Infrastructure to track locations..."
Markus Armbruster [Mon, 22 Mar 2010 09:29:02 +0000 (10:29 +0100)]
error: Trim includes after "Infrastructure to track locations..."

Missed in commit 827b0813.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoerror: Trim includes in qerror.c
Markus Armbruster [Mon, 22 Mar 2010 09:29:01 +0000 (10:29 +0100)]
error: Trim includes in qerror.c

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoerror: Trim includes after "Move qemu_error & friends..."
Markus Armbruster [Mon, 22 Mar 2010 09:29:00 +0000 (10:29 +0100)]
error: Trim includes after "Move qemu_error & friends..."

Missed in commit 2f792016.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agolinux-user: Use RLIMIT_STACK for default stack size.
Richard Henderson [Fri, 19 Mar 2010 21:21:13 +0000 (14:21 -0700)]
linux-user: Use RLIMIT_STACK for default stack size.

The current default stack limit of 512kB is far too small; a fair
number of gcc testsuite failures (for all guests) are directly
attributable to this.  Using the -s option in every invocation of
the emulator is annoying to say the least.

A reasonable compromise seems to be to honor the system rlimit.
At least on two Linux distributions, this is set to 8MB and 10MB
respectively.  If the system does not limit the stack, then we're
no worse off than before.

At the same time, rename the variable from x86_stack_size and
change the ultimate fallback size from 512kB to 8MB.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotarget-arm: disable PAGE_EXEC for XN pages
Rabin Vincent [Fri, 19 Mar 2010 20:58:03 +0000 (02:28 +0530)]
target-arm: disable PAGE_EXEC for XN pages

Don't set PAGE_EXEC for XN pages, to avoid a bypass of XN protection
checking if the page is already in the TLB.

Signed-off-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agofix race between timer firing vs. alarm_timer->pending = 0
Paolo Bonzini [Fri, 19 Mar 2010 10:30:35 +0000 (11:30 +0100)]
fix race between timer firing vs. alarm_timer->pending = 0

The period for Win32 timers is very short and always the same
independent of dynticks, so it's possible that the timer fires
before qemu_run_all_timers has reset alarm_timer->pending to zero.
Reset alarm_timer->pending before rearming.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoqemu-io: Fix return value handling of bdrv_open
Ryota Ozaki [Sat, 20 Mar 2010 07:08:38 +0000 (16:08 +0900)]
qemu-io: Fix return value handling of bdrv_open

bdrv_open may return -errno so we have to check
if the return value is '< 0', not '== -1'.

Signed-off-by: Ryota Ozaki <ozaki.ryota@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>