sdk/emulator/qemu.git
14 years agosparc64: fix 32bit load sign extension
Igor V. Kovalenko [Tue, 1 Jun 2010 20:12:32 +0000 (00:12 +0400)]
sparc64: fix 32bit load sign extension

- change return type of ldl_* to uint32_t to prevent unwanted sign extension
  visible in sparc64 load alternate address space methods
- note this change makes ldl_* softmmu implementations match ldl_phys one
Signed-off-by: Igor V. Kovalenko <igor.v.kovalenko@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agosparc64: fix missing address masking v1
Igor V. Kovalenko [Wed, 2 Jun 2010 19:38:45 +0000 (23:38 +0400)]
sparc64: fix missing address masking v1

- address masking for ldqf and stqf insns
- address masking for lddf and stdf insns
- address masking for translating ASI (Ultrasparc IIi)
v0->v1:
- move arch-specific code to helpers and drop more ifdefs at call sites
  using new helper asi_address_mask()
- change user emulation to use asi_address_mask()

Signed-off-by: Igor V. Kovalenko <igor.v.kovalenko@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agosparc64: fix tag access register on mmu traps
Igor V. Kovalenko [Tue, 1 Jun 2010 20:12:22 +0000 (00:12 +0400)]
sparc64: fix tag access register on mmu traps

- set mmu tag access register on FAULT and PROT traps as well

Signed-off-by: Igor V. Kovalenko <igor.v.kovalenko@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agovnc: add missing target for vnc-encodings-*.o
Corentin Chary [Tue, 1 Jun 2010 21:05:44 +0000 (23:05 +0200)]
vnc: add missing target for vnc-encodings-*.o

vnc-encodings-*.c dependencies where missing.

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: tight: don't forget the third color
Corentin Chary [Tue, 1 Jun 2010 21:05:43 +0000 (23:05 +0200)]
vnc: tight: don't forget the third color

While couting color, if the third color was only present one
time it wasn't added to the palette.

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: tight: don't forget last pixel in tight_encode_indexed_rect
Corentin Chary [Tue, 1 Jun 2010 21:05:42 +0000 (23:05 +0200)]
vnc: tight: don't forget last pixel in tight_encode_indexed_rect

A simple patch would have been to just remove count -= 1, but this
one also replace the while (count--) with a for(i = 0; i < count; i++)
which I believe is more easy to understand.

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoMonitor: Drop QMP documentation from code
Luiz Capitulino [Mon, 31 May 2010 17:43:32 +0000 (14:43 -0300)]
Monitor: Drop QMP documentation from code

Previous commit added QMP documentation to the qemu-monitor.hx
file, it's is a copy of this information.

While it's good to keep it near code, maintaining two copies of
the same information is too hard and has little benefit as we
don't expect client writers to consult the code to find how to
use a QMP command.

Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoQMP: Introduce commands documentation
Jan Kiszka [Mon, 31 May 2010 17:43:31 +0000 (14:43 -0300)]
QMP: Introduce commands documentation

One of the most important missing feature in QMP today is its
supported commands documentation.

The plan is to make it part of self-description support, however
self-description is a big task we have been postponing for a
long time now and still don't know when it's going to be done.

In order not to compromise QMP adoption and make users' life easier,
this commit adds a simple text documentation which fully describes
all QMP supported commands.

This is not ideal for a number of reasons (harder to maintain,
text-only, etc) but does improve the current situation. To avoid at
least divering from the user monitor help and texi snippets, QMP bits
are also maintained inside qemu-monitor.hx, and hxtool is extended to
generate a single text file from them.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agomonitor: Reorder info documentation
Jan Kiszka [Mon, 31 May 2010 17:43:30 +0000 (14:43 -0300)]
monitor: Reorder info documentation

Push the doc fragments for the info command to the end of
qemu-monitor.hx. This helps to establish a proper layout in the upcoming
QMP documentation.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoMerge remote branch 'mst/for_anthony' into HEAD
Anthony Liguori [Tue, 1 Jun 2010 18:34:20 +0000 (13:34 -0500)]
Merge remote branch 'mst/for_anthony' into HEAD

14 years agovnc: move size-changed check into the vnc_desktop_resize function.
Gerd Hoffmann [Tue, 25 May 2010 16:25:20 +0000 (18:25 +0200)]
vnc: move size-changed check into the vnc_desktop_resize function.

This make sure we send a desktop resize message only in case we actually
have to, using the new variables which track the clients desktop size.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: don't send invalid screen updates.
Gerd Hoffmann [Tue, 25 May 2010 16:25:19 +0000 (18:25 +0200)]
vnc: don't send invalid screen updates.

Don't send updates for screen areas which are outside the clients
desktop.  May happed with vnc clients which don't support the desktop
resize message.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: keep track of client desktop size
Gerd Hoffmann [Tue, 25 May 2010 16:25:18 +0000 (18:25 +0200)]
vnc: keep track of client desktop size

Add two new variables to keep track of the vnc clients desktop size.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: send desktopresize event as reply to set encodings
Gerd Hoffmann [Tue, 25 May 2010 16:25:17 +0000 (18:25 +0200)]
vnc: send desktopresize event as reply to set encodings

In case the desktop did resize while the vnc connection setup was still
in progress the client isn't informed about it.  Send a desktop resize
event as soon as the client told us it can handle deskop resize via set
encodings message to make sure the client us up to date.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: factor out vnc_desktop_resize()
Gerd Hoffmann [Tue, 25 May 2010 16:25:16 +0000 (18:25 +0200)]
vnc: factor out vnc_desktop_resize()

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoAdd support for depth 15 to qemu_default_pixelformat()
Gerd Hoffmann [Fri, 21 May 2010 09:59:14 +0000 (11:59 +0200)]
Add support for depth 15 to qemu_default_pixelformat()

Makes qemu_default_pixelformat(15) return pixelformat filled for 15 bit
color depth (16 bpp, 5 bits for red,green,blue each, 1 bit unused).

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agocheck for active_console before using it
Gerd Hoffmann [Thu, 20 May 2010 13:23:06 +0000 (15:23 +0200)]
check for active_console before using it

Other vga_hw_* functions do the same.
Fixes a segmentation fault.  Trigger: boot with -nodefaults,
then connect via vnc.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoAdd dependency of JSON unit tests on config-host.h
Jan Kiszka [Thu, 20 May 2010 07:18:52 +0000 (09:18 +0200)]
Add dependency of JSON unit tests on config-host.h

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agomigration: Fix calculation of bytes_transferred
Pierre Riteau [Wed, 12 May 2010 13:12:44 +0000 (15:12 +0200)]
migration: Fix calculation of bytes_transferred

When a page with all identical bytes is transferred, it is counted
as a full page (TARGET_PAGE_SIZE) although only one byte is actually
sent. Fix this by changing ram_save_block() to return the number of
bytes sent instead of a boolean value. This makes bandwidth
estimation, and consequently downtime estimation, more precise.

Signed-off-by: Pierre Riteau <Pierre.Riteau@irisa.fr>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovirtio-serial-bus: fix ports_map allocation on init
Alon Levy [Wed, 19 May 2010 09:31:38 +0000 (15:01 +0530)]
virtio-serial-bus: fix ports_map allocation on init

Fix for too small allocation to ports_map

Signed-off-by: Alon Levy <alevy@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoresent: x86/cpuid: Add kvm32 CPU model
Andre Przywara [Fri, 21 May 2010 07:50:51 +0000 (09:50 +0200)]
resent: x86/cpuid: Add kvm32 CPU model

Create a kvm32 CPU model that describes a least common denominator
for KVM capable guest CPUs. Useful for migration purposes.

Signed-off-by: Andre Przywara <andre.przywara@amd.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: tight: add palette encoding
Corentin Chary [Wed, 19 May 2010 07:24:12 +0000 (09:24 +0200)]
vnc: tight: add palette encoding

Add palette tight encoding. Palette encoding will try to count the number
of colors for a given rectangle, and if this number is low enough
(< 256) it will send the palette + the rectangle with indexed colors.

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: add support for tight fill encoding
Corentin Chary [Wed, 19 May 2010 07:24:11 +0000 (09:24 +0200)]
vnc: add support for tight fill encoding

Fill encoding detects rectangles using only one color and send only
one pixel value.

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: add basic tight support
Corentin Chary [Wed, 19 May 2010 07:24:10 +0000 (09:24 +0200)]
vnc: add basic tight support

Add support for tight encoding [1]. This patch only add support
for "basic" tight compression without any filter.

[1] http://tigervnc.org/cgi-bin/rfbproto#tight-encoding.

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: return the number of rectangles
Corentin Chary [Wed, 19 May 2010 07:24:09 +0000 (09:24 +0200)]
vnc: return the number of rectangles

Some encodings like tight supports tiling (spliting in
multiple sub-rectangles). So we needed a way to tell
vnc_update_client() how much rectangles are in the buffer.

zlib, raw and hextile always send a full rectangle.

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: remove a memory leak in zlib
Corentin Chary [Wed, 19 May 2010 07:24:08 +0000 (09:24 +0200)]
vnc: remove a memory leak in zlib

Makes sure we free all ressources used in zlib encoding (zlib stream
and buffer).

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: add buffer_free()
Corentin Chary [Wed, 19 May 2010 07:24:07 +0000 (09:24 +0200)]
vnc: add buffer_free()

Add a buffer_free() helper to free vnc buffers and
remove some duplicated code in vnc_disconnect_finish().

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: don't clear zlib stream on set_encoding
Corentin Chary [Wed, 19 May 2010 07:24:06 +0000 (09:24 +0200)]
vnc: don't clear zlib stream on set_encoding

On init, values are already NULL, but we shouldn't try
to reset them each time a client send a set encoding
command because this break everything. For example,
libvncclient re-send a set encoding command if the
framebuffer is resized.

This fix framebuffer resizing for zlib encoding.

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: adjust compression zstream level
Corentin Chary [Wed, 19 May 2010 07:24:05 +0000 (09:24 +0200)]
vnc: adjust compression zstream level

Adjust zlib compression level if needed by calling deflateParams.

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: only use a single zlib stream
Corentin Chary [Wed, 19 May 2010 07:24:04 +0000 (09:24 +0200)]
vnc: only use a single zlib stream

According to http://tigervnc.org/cgi-bin/rfbproto#zlib-encoding
zlib encoding only uses a single stream. Current implementation defines
4 streams but only uses the first one. Remove them and only use a single
stream.

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: really call zlib if we want zlib
Corentin Chary [Wed, 19 May 2010 07:24:03 +0000 (09:24 +0200)]
vnc: really call zlib if we want zlib

send_framebuffer_update() was calling hextile instead of zlib
since commit 70a4568fe0c5a64adaa3da5030b7109e5199e692.

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agovnc: explain why set_encodings loop is reversed
Corentin Chary [Wed, 19 May 2010 07:24:02 +0000 (09:24 +0200)]
vnc: explain why set_encodings loop is reversed

Add a small comment to explain why we need to start from
the end of the array to set the right prefered encoding.

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoRevert "vnc: set the right prefered encoding"
Corentin Chary [Wed, 19 May 2010 07:24:01 +0000 (09:24 +0200)]
Revert "vnc: set the right prefered encoding"

This patch was wrong, because the loop was already reversed,
so the first encoding was correctly set at the end of the loopp.

This reverts commit 14eb8b6829ad9dee7035de729e083844a425f274.

Signed-off-by: Corentin Chary <corentincj@iksaif.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoioapic: improve debugging
Blue Swirl [Mon, 31 May 2010 18:59:45 +0000 (18:59 +0000)]
ioapic: improve debugging

Add a DPRINTF macro, use it also to see irq deliveries.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoFix leul_to_cpu on big endian hosts
Alexander Graf [Tue, 11 May 2010 15:38:17 +0000 (17:38 +0200)]
Fix leul_to_cpu on big endian hosts

Commit 213acd2e introduced leul_to_cpu with a special code path for big endian
hosts. Unfortunately that code used preprocessor magic that didn't work.

This patch replaces the explicit ##s by glue() which is proven to work reliably,
enabling me to compile qemu on ppc again.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoarm: fix arm kernel boot for non zero start addr
Lars Munch [Sat, 8 May 2010 20:43:35 +0000 (22:43 +0200)]
arm: fix arm kernel boot for non zero start addr

Booting an arm kernel has been broken a while when booting from non zero start
address. This is due to the order of events: board init loads the kernel and
sets register 15 to the start address and then qemu_system_reset reset the cpu
making register 15 zero again.

This patch fixes the usage of the register 15 start address trick in
combination with arm_load_kernel.

Signed-off-by: Lars Munch <lars@segv.dk>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotarget-ppc: remove useless line
Thomas Monjalon [Fri, 28 May 2010 19:00:45 +0000 (21:00 +0200)]
target-ppc: remove useless line

This line was a bit clear.
The next lines set or reset this bit (LE) depending of another bit (ILE).
So the first line is useless.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotarget-ppc: fix RFI by clearing some bits of MSR
Thomas Monjalon [Fri, 28 May 2010 19:07:32 +0000 (21:07 +0200)]
target-ppc: fix RFI by clearing some bits of MSR

Since commit 2ada0ed, "Return From Interrupt" is broken for PPC processors
because some interrupt specifics bits of SRR1 are copied to MSR.

SRR1 is a save of MSR during interrupt.
During RFI, MSR must be restored from SRR1.
But some bits of SRR1 are interrupt-specific and are not used for MSR saving.

This is the specification (ISA 2.06) at chapter 6.4.3 (Interrupt Processing):
"2. Bits 33:36 and 42:47 of SRR1 or HSRR1 are loaded with information specific
    to the interrupt type.
 3. Bits 0:32, 37:41, and 48:63 of SRR1 or HSRR1 are loaded with a copy of the
    corresponding bits of the MSR."

Below is a representation of MSR bits which are not saved:
0:15 16:31 32  33:36    37:41      42:47     48:63
——— | ——— | — X X X X — — — — — X X X X X X | ————
0000 0000 |    7   |   8   |   3   |   F    | 0000

History:
In the initial Qemu implementation (e1833e1), the mask 0x783F0000 was used for
saving MSR in SRR1. But all the bits 32:47 were cleared during RFI restoring.
This was wrong. The commit 2ada0ed explains that this breaks Altivec.
Indeed, bit 38 (for Altivec support) must be saved and restored.
The change of 2ada0ed was to restore all the bits of SRR1 to MSR.
But it's also wrong.

Explanation:
As an example, let's see what's happening after a TLB miss.
According to the e300 manual (E300CORERM table 5-6), the TLB miss interrupts
set the bits 44-47 for KEY, I/D, WAY and S/L. These bits are specifics to the
interrupt and must not be copied into MSR at the end of the interrupt.
With the current implementation, a TLB miss overwrite bits POW, TGPR and ILE.

Fix:
It shouldn't be needed to filter-out bits on MSR saving when interrupt occurs.
Specific bits overwrite MSR ones in SRR1.
But at the end of interrupt (RFI), specifics bits must be cleared before
restoring MSR from SRR1. The mask 0x783F0000 apply here.

Discussion:
The bits of the mask 0x783F0000 are cleared after an interrupt.
I cannot find a specification which talks about this
but I assume it is the truth since Linux can run this way.
Maybe it's not perfect but it's better (works for e300).

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agopci-hotplug: make them aware of pci domain.
Isaku Yamahata [Fri, 28 May 2010 09:30:46 +0000 (18:30 +0900)]
pci-hotplug: make them aware of pci domain.

add helper function which converts root bus to pci domain.
make them aware of pci domain.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agomsix: remove duplicated defines.
Isaku Yamahata [Thu, 27 May 2010 05:40:37 +0000 (14:40 +0900)]
msix: remove duplicated defines.

remove defines which are already defined in pci_regs.h

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agopci.h: remove unused constants.
Isaku Yamahata [Thu, 27 May 2010 05:40:02 +0000 (14:40 +0900)]
pci.h: remove unused constants.

So remove unused constants,
PCI_STATUS_RESERVED_MASK_LO, PCI_STATUS_RESERVED_MASK_HI,
PCI_COMMAND_RESERVED, PCI_COMMAND_RESERVED_MASK_HI.
They were used once, but they aren't used anymore.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agopci: add const to pci_is_express(), pci_config_size().
Isaku Yamahata [Thu, 27 May 2010 05:42:37 +0000 (14:42 +0900)]
pci: add const to pci_is_express(), pci_config_size().

add const to pci_is_express(), pci_config_size().

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agopci: clean up of pci_set_default_subsystem_id().
Isaku Yamahata [Thu, 27 May 2010 05:42:06 +0000 (14:42 +0900)]
pci: clean up of pci_set_default_subsystem_id().

Use pci accessor function.
don't return value because it always return 0 and
the caller doesn't check the return value.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agomc146818rtc: improve debugging
Blue Swirl [Sun, 30 May 2010 19:20:07 +0000 (19:20 +0000)]
mc146818rtc: improve debugging

Add a separate flag for debugging coalesced interrupts.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agosparc32 SuperSPARC MMU Breakpoint Action register (SS-20 OBP fix)
Artyom Tarasenko [Sat, 29 May 2010 20:48:25 +0000 (22:48 +0200)]
sparc32 SuperSPARC MMU Breakpoint Action register (SS-20 OBP fix)

SuperSPARC MMU Breakpoint Action register is used by OBP at boot

The patch allows booting Solaris and some other OS with
SPARCStation-20 OBP.

Signed-off-by: Artyom Tarasenko <atar4qemu@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agopc: improve debugging
Blue Swirl [Sat, 29 May 2010 20:23:49 +0000 (20:23 +0000)]
pc: improve debugging

Add a DPRINTF macro and use it for ISA and PIC interrupts.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoapic: improve debugging
Blue Swirl [Sat, 29 May 2010 20:23:48 +0000 (20:23 +0000)]
apic: improve debugging

Add a DPRINTF macro. Use TARGET_FMT_plx for printing target_phys_addr_t
items. Add a separate flag for debugging coalescing interrupts.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoi8259: improve debugging
Blue Swirl [Sat, 29 May 2010 20:23:19 +0000 (20:23 +0000)]
i8259: improve debugging

Add a DPRINTF macro.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agosparc64: fix user emulator build
Blue Swirl [Sat, 29 May 2010 10:20:44 +0000 (10:20 +0000)]
sparc64: fix user emulator build

Accesses with _nucleus prefix are not available when building user
emulators:
  CC    sparc64-linux-user/op_helper.o
cc1: warnings being treated as errors
/src/qemu/target-sparc/op_helper.c: In function 'helper_ldda_asi':
/src/qemu/target-sparc/op_helper.c:3063: error: implicit declaration of function 'ldq_nucleus'

Avoid making such accesses. Fixes breakage by 54a3c0f032abf93d1cedc75456af75401ac10360.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoPad iommu with an empty slot (necessary for SunOS 4.1.4)
Artyom Tarasenko [Sat, 8 May 2010 14:10:14 +0000 (16:10 +0200)]
Pad iommu with an empty slot (necessary for SunOS 4.1.4)

On the real hardware (SS-5, LX) the MMU is not padded, but aliased.
Software shouldn't use aliased addresses, neither should it crash
when it uses (on the real hardware it wouldn't). Using empty_slot
instead of aliasing can help with debugging such accesses.

Signed-off-by: Artyom Tarasenko <atar4qemu@googlemail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agosparc64: fix 128-bit atomic load from nucleus context v1
Igor V. Kovalenko [Fri, 28 May 2010 21:05:46 +0000 (01:05 +0400)]
sparc64: fix 128-bit atomic load from nucleus context v1

- change 128-bit atomic loads to reference nucleus context
v0->v1: dropped disassembler change
Signed-off-by: Igor V. Kovalenko <igor.v.kovalenko@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoPre-allocate guest address space
Paul Brook [Sat, 29 May 2010 01:27:35 +0000 (02:27 +0100)]
Pre-allocate guest address space

Allow pre-allocation of the guest virtual address space in usermode emulation.

Signed-off-by: Paul Brook <paul@codesourcery.com>
14 years agoalpha-linux-user: Fill in SI_CODE for SIGSEGV.
Richard Henderson [Fri, 21 May 2010 17:03:33 +0000 (10:03 -0700)]
alpha-linux-user: Fill in SI_CODE for SIGSEGV.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agolinux-user: Use qemu-malloc.c.
Richard Henderson [Fri, 21 May 2010 17:37:52 +0000 (10:37 -0700)]
linux-user: Use qemu-malloc.c.

Since we're no longer setting PAGE_RESERVED, there's no need to
implement qemu_malloc via mmap.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoUse calloc in qemu_mallocz.
Richard Henderson [Fri, 21 May 2010 17:37:51 +0000 (10:37 -0700)]
Use calloc in qemu_mallocz.

Avoids the memset if the allocator has gotten new zeroed
storage from the operating system.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agolinux-user: Fix compilation with DEBUG defined
Riccardo Magliocchetti [Wed, 19 May 2010 19:49:21 +0000 (21:49 +0200)]
linux-user: Fix compilation with DEBUG defined

Signed-off-by: Riccardo Magliocchetti <riccardo.magliocchetti@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoFix hw/gt64xxx.c compilation with DEBUG defined
Riccardo Magliocchetti [Wed, 19 May 2010 16:49:30 +0000 (18:49 +0200)]
Fix hw/gt64xxx.c compilation with DEBUG defined

Use TARGET_FMT_plx as format placeholder for target_phys_addr_t

Signed-off-by: Riccardo Magliocchetti <riccardo.magliocchetti@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoFix typo in balloon help
Riccardo Magliocchetti [Wed, 19 May 2010 16:49:28 +0000 (18:49 +0200)]
Fix typo in balloon help

Fix launchpad #563883

Signed-off-by: Riccardo Magliocchetti <riccardo.magliocchetti@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agomusicpal: Drop redundant reset call
Jan Kiszka [Thu, 13 May 2010 13:09:58 +0000 (15:09 +0200)]
musicpal: Drop redundant reset call

Reset is now triggered after init, no need for explicit calls anymore.

Signed-off-by: Jan Kiszka <jan.kiszka@web.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agodoc: Update monitor info subcommands
Stefan Hajnoczi [Wed, 12 May 2010 14:10:56 +0000 (15:10 +0100)]
doc: Update monitor info subcommands

The "info blockstats" documentation was copy-pasted as "info block"
instead of "info blockstats".  The documentation for "commands", "jit",
"numa", "qdm", and "roms" is missing.  This patch resolves these issues
in qemu-monitor.hx.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoFix overflow in i440fx_init()
Avi Kivity [Sun, 9 May 2010 11:51:13 +0000 (14:51 +0300)]
Fix overflow in i440fx_init()

The ram_size parameter can be larger than an int, so it may be truncated.

Fix by using the correct type.

Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agoarm: prevent coprocessor IO reset
Lars Munch [Sat, 8 May 2010 20:42:43 +0000 (22:42 +0200)]
arm: prevent coprocessor IO reset

This prevent coprocessor IO structure from being reset on cpu reset. This was
a problem for PXA which uses coprocessor 6 and 14.

Signed-off-by: Lars Munch <lars@segv.dk>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotcg: Use INDEX_op_qemu_ld32 for 32-bit results.
Richard Henderson [Mon, 3 May 2010 23:30:48 +0000 (16:30 -0700)]
tcg: Use INDEX_op_qemu_ld32 for 32-bit results.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agovhost_net.c: v2 Fix build failure introduced by 0bfcd599e3f5c5679cc7d0165a0a1822e2f60de2
Jes Sorensen [Thu, 27 May 2010 12:26:05 +0000 (14:26 +0200)]
vhost_net.c: v2 Fix build failure introduced by 0bfcd599e3f5c5679cc7d0165a0a1822e2f60de2

Fix build failure introduced by 0bfcd599e3f5c5679cc7d0165a0a1822e2f60de2

The format statement expects unsigned long on x86_64, but receives
unsigned long long, so gcc exits with an error.

Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoMake cache=unsafe the default for -snapshot
Alexander Graf [Wed, 26 May 2010 19:04:32 +0000 (21:04 +0200)]
Make cache=unsafe the default for -snapshot

When using -snapshot we don't care about data integrity of the cow file
at all, so let's disable flushing there and squeeze out the last drop
of performance we could possibly get.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agosparc64: clean up pci bridge map
Igor V. Kovalenko [Tue, 25 May 2010 12:09:03 +0000 (16:09 +0400)]
sparc64: clean up pci bridge map

- remove unused host state and store pci bus pointer only
- do not map host state access into unused 1fe.10000000 range
- reorder pci region registration
- assign pci i/o region to isa_mem_base

Signed-off-by: Igor V. Kovalenko <igor.v.kovalenko@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agosparc64: rename sun4u cpu to Ultrasparc IIi
Igor V. Kovalenko [Tue, 25 May 2010 12:08:57 +0000 (16:08 +0400)]
sparc64: rename sun4u cpu to Ultrasparc IIi

Signed-off-by: Igor V. Kovalenko <igor.v.kovalenko@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoAdd cache=unsafe parameter to -drive
Alexander Graf [Wed, 26 May 2010 15:51:49 +0000 (17:51 +0200)]
Add cache=unsafe parameter to -drive

Usually the guest can tell the host to flush data to disk. In some cases we
don't want to flush though, but try to keep everything in cache.

So let's add a new cache value to -drive that allows us to set the cache
policy to most aggressive, disabling flushes. We call this mode "unsafe",
as guest data is not guaranteed to survive host crashes anymore.

This patch also adds a noop function for aio, so we can do nothing in AIO
fashion.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agovnc: rich cursor support.
Gerd Hoffmann [Fri, 21 May 2010 09:54:34 +0000 (11:54 +0200)]
vnc: rich cursor support.

Uses VNC_ENCODING_RICH_CURSOR.  Adding XCURSOR support should be
possible without much trouble.  Shouldn't be needed though as
RICH_CURSOR is a superset of XCURSOR.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agouse new cursor struct + functions for vmware vga and sdl.
Gerd Hoffmann [Fri, 21 May 2010 09:54:33 +0000 (11:54 +0200)]
use new cursor struct + functions for vmware vga and sdl.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agocursor: add cursor functions.
Gerd Hoffmann [Fri, 21 May 2010 09:54:32 +0000 (11:54 +0200)]
cursor: add cursor functions.

Add a new cursor type to console.h and a bunch of functions to
deal with cursors the (new) cursor.c file.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoFix error handling in qemu_read_config_file
Kevin Wolf [Mon, 17 May 2010 08:36:47 +0000 (10:36 +0200)]
Fix error handling in qemu_read_config_file

We need to close the file even in error case. While at it, make the callers
catch all kind of errors. ENOENT is allowed for default config files, they
are optional.

Reported-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoVirtio-net: Replace the hardcode 6 with defined ETN_ALEN
Amos Kong [Sun, 23 May 2010 02:49:10 +0000 (10:49 +0800)]
Virtio-net: Replace the hardcode 6 with defined ETN_ALEN

hw/virtio-net.h:
    #define ETH_ALEN    6
ETH_ALEN was defined by commit 7967406801aa897fae83caad3278ac85a342adaa

Signed-off-by: Amos Kong <akong@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agosdl: Do not disable screensaver by default
Jan Kiszka [Sun, 23 May 2010 08:29:34 +0000 (10:29 +0200)]
sdl: Do not disable screensaver by default

Unless we are running in full-screen mode, QEMU's SDL window should not
disable the host's screensaver. The user can still change this behaviour
by setting the environment variable SDL_VIDEO_ALLOW_SCREENSAVER as
desired.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoMerge remote branch 'kwolf/for-anthony' into staging
Anthony Liguori [Mon, 24 May 2010 15:53:41 +0000 (10:53 -0500)]
Merge remote branch 'kwolf/for-anthony' into staging

14 years agoMerge remote branch 'qmp/for-anthony' into staging
Anthony Liguori [Mon, 24 May 2010 14:00:24 +0000 (09:00 -0500)]
Merge remote branch 'qmp/for-anthony' into staging

14 years agomicroblaze: Handle new elf mach nr for sysemu.
Edgar E. Iglesias [Mon, 24 May 2010 09:14:04 +0000 (11:14 +0200)]
microblaze: Handle new elf mach nr for sysemu.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
14 years agolsi: Fix value overflow in request tag processing
Jan Kiszka [Fri, 21 May 2010 16:44:59 +0000 (18:44 +0200)]
lsi: Fix value overflow in request tag processing

This fixes a mismerge of 64d564094cac5f72eeaeb950c442b773a00d3586 (wrong
patch version): We need to mask the tag value properly to obtain its
device ID.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agosparc64: flush translations on mmu context change
Igor V. Kovalenko [Sat, 22 May 2010 10:52:40 +0000 (14:52 +0400)]
sparc64: flush translations on mmu context change

- two pairs of softmmu indexes bind softmmu tlb to cpu tlb in fault handlers
  using value of DMMU primary and secondary context registers, so we need to
  flush softmmu translations when context registers are changed

Signed-off-by: Igor V. Kovalenko <igor.v.kovalenko@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agosparc64: fix mmu context at trap levels above zero
Igor V. Kovalenko [Sat, 22 May 2010 10:52:35 +0000 (14:52 +0400)]
sparc64: fix mmu context at trap levels above zero

- cpu_mmu_index return MMU_NUCLEUS_IDX if trap level is not zero
- cpu_get_tb_cpu_state: store trap level and primary context in flags
  this allows to restart code translation when address translation is changed
- stop translation block after writing to pstate and tl registers
- stop translation block after writing to alternate space
  this can be optimized to stop only if address translation can be changed
  by write operation (e.g. by comparing with MMU ASI values)

Signed-off-by: Igor V. Kovalenko <igor.v.kovalenko@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agosparc64: fix dump_mmu to look for global bit in tte value instead of tag
Igor V. Kovalenko [Sat, 22 May 2010 10:52:29 +0000 (14:52 +0400)]
sparc64: fix dump_mmu to look for global bit in tte value instead of tag

Signed-off-by: Igor V. Kovalenko <igor.v.kovalenko@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agosparc64: fix pstate privilege bits
Igor V. Kovalenko [Sat, 22 May 2010 10:52:24 +0000 (14:52 +0400)]
sparc64: fix pstate privilege bits

- refactor code to handle hpstate only if available for current cpu
- conditionally set hypervisor bit in hpstate register
- reorder softmmu indices so user accessable ones go first, translation context
  macros supervisor() and hypervisor() adjusted as well
- disable sparcv8 registers for TARGET_SPARC64
- fix cpu_mmu_index to use sparcv9 bits only

Signed-off-by: Igor V. Kovalenko <igor.v.kovalenko@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agosparc64: generate data access exception on RW violation
Igor V. Kovalenko [Sat, 22 May 2010 10:52:19 +0000 (14:52 +0400)]
sparc64: generate data access exception on RW violation

- separate PRIV and PROT handling
- DPRINTF_MMU macro to clean up debug code
- dump mmu_idx, trap level and mmu context registers
  along with address translation values

Signed-off-by: Igor V. Kovalenko <igor.v.kovalenko@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agosparc32 protect read-only bits in DMA CSR registers
Artyom Tarasenko [Sat, 22 May 2010 08:38:56 +0000 (10:38 +0200)]
sparc32 protect read-only bits in DMA CSR registers

On a real hardware changing read-only bits has no effect
Use a mask common for SCSI and Ethernet registers. The crucial
bit is DMA_INTR, because setting or clearing it may produce
spurious interrupts.

This patch allows booting Solaris 2.3

Signed-off-by: Artyom Tarasenko <atar4qemu@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agovmstate: fix breakage by 7e72abc382b700a72549e8147bdea413534eeedc
TeLeMan [Wed, 12 May 2010 02:28:51 +0000 (10:28 +0800)]
vmstate: fix breakage by 7e72abc382b700a72549e8147bdea413534eeedc

cirrus_post_load() will be executed twice when loading vm states and then the
wrong physical memory will be registered. This issue may lead to crash qemu.

Signed-off-by: TeLeMan <geleman@gmail.com>
Acked-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoFix -device help and documentation
Markus Armbruster [Tue, 11 May 2010 12:02:31 +0000 (14:02 +0200)]
Fix -device help and documentation

Commit 6616b2ad reverted commit 40ea285c.  Looks like a mismerge to
me.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoFix tarbin Makefile rule
Stuart Brady [Wed, 12 May 2010 19:42:04 +0000 (20:42 +0100)]
Fix tarbin Makefile rule

The 'tarbin' Makefile rule doesn't include qemu-system-sparc64, but
should do, now that sparc64-softmmu is in the default target list.

The rule attempts to tar up binaries that were not built if a target
list was passed to the configure script -- in which case, it will
either fail, or otherwise include binaries from previous builds.

Fix both problems once and for all by building a list of binaries to
include in the tarball, using the list of targets to be built.

Signed-off-by: Stuart Brady <sdb@zubnet.me.uk>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agohxtool: Add syntax error detection
Jan Kiszka [Thu, 20 May 2010 07:16:33 +0000 (09:16 +0200)]
hxtool: Add syntax error detection

Add basic imbalance detection for STEXT/ETEXI.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoFix TEXI section mark imbalance in qemu-img-cmd.hx
Jan Kiszka [Thu, 20 May 2010 07:16:30 +0000 (09:16 +0200)]
Fix TEXI section mark imbalance in qemu-img-cmd.hx

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoPut dependency files in proper subdir
Jan Kiszka [Sat, 15 May 2010 11:03:28 +0000 (13:03 +0200)]
Put dependency files in proper subdir

This seems to resolve subtle breakages of our build system:

Dependency files generated for targets like 'dir/foo.o' were saved as
'foo.d'. Now, if there was also a target 'foo.o', one of the dependency
file was overwritten. Concrete example: libhw*/macio.o vs.
libhw*/ide/macio.o. And this often left a segfaulting build result
behind when changing the "wrong" data structures".

Fix it by generating proper 'dir/foo.d'.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoClean libhw subdirs as well
Jan Kiszka [Sat, 15 May 2010 11:03:30 +0000 (13:03 +0200)]
Clean libhw subdirs as well

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoFix %lld or %llx printf format use
Blue Swirl [Sat, 22 May 2010 08:02:12 +0000 (08:02 +0000)]
Fix %lld or %llx printf format use

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoCompile dma only once
Blue Swirl [Sat, 22 May 2010 08:00:52 +0000 (08:00 +0000)]
Compile dma only once

Use a qemu_irq to request CPU exit.

7 compilations less for the full build.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agopckbd: improve debugging
Blue Swirl [Sat, 22 May 2010 07:59:06 +0000 (07:59 +0000)]
pckbd: improve debugging

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoCompile pckbd only once
Blue Swirl [Sat, 22 May 2010 07:59:01 +0000 (07:59 +0000)]
Compile pckbd only once

Use a qemu_irq to indicate A20 line changes. Move I/O port 92
to pckbd.c.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agolinux-user: do not warn for missing pselect6
Riku Voipio [Fri, 7 May 2010 12:28:05 +0000 (12:28 +0000)]
linux-user: do not warn for missing pselect6

Libc will fallback gracefully if pselect6 is not available. Thus put
pselect6 to nowarn until the atomicity issues of the original pselect6
patch are dealt with.

Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
Cc: Michael Casadevall <mcasadevall@ubuntu.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotcg-ia64: Load GUEST_BASE into a register.
Richard Henderson [Thu, 6 May 2010 15:50:44 +0000 (08:50 -0700)]
tcg-ia64: Load GUEST_BASE into a register.

Saves one bundle per memory operation.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotcg-ia64: Fix some register usage issues.
Richard Henderson [Thu, 6 May 2010 15:50:43 +0000 (08:50 -0700)]
tcg-ia64: Fix some register usage issues.

(1) The output registers were not marked call-clobbered, even though
    they can be modified by called functions.
(2) The thread pointer was not marked reserved.
(3) R4-R6 are call-saved, but not saved by the prologue.  Rather than
    save them, mark them reserved so that we don't use them.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotcg-hppa: Load GUEST_BASE as an immediate.
Richard Henderson [Thu, 6 May 2010 15:50:42 +0000 (08:50 -0700)]
tcg-hppa: Load GUEST_BASE as an immediate.

Now that the prologue is generated after GUEST_BASE is fixed,
we can load it as an immediate, and also avoid reserving the
register if it isn't necessary.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
14 years agotcg: Initialize the prologue after GUEST_BASE is fixed.
Richard Henderson [Thu, 6 May 2010 15:50:41 +0000 (08:50 -0700)]
tcg: Initialize the prologue after GUEST_BASE is fixed.

This will allow backends to make intelligent choices about how
to implement GUEST_BASE.

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