Current implementation of thumb mul instruction is implemented as a
32x32->64 multiply which then uses only 32 least significant bits of
the result. Replace that with a simple 32x32->32 multiply.
Juha Riihimäki [Thu, 22 Oct 2009 12:17:36 +0000 (15:17 +0300)]
target-arm: cleanup internal resource leaks
Revised patch for getting rid of tcg temporary variable leaks in
target-arm/translate.c. This version also includes the leak patch for
gen_set_cpsr macro, now converted as a static inline function, which I
sent earlier as a separate patch on top of this patch.
Andre Przywara [Fri, 23 Oct 2009 11:44:31 +0000 (13:44 +0200)]
target-i386: implement lzcnt emulation
lzcnt is a AMD Phenom/Barcelona added instruction returning the
number of leading zero bits in a word.
As this is similar to the "bsr" instruction, reuse the existing
code. There need to be some more changes, though, as lzcnt always
returns a valid value (in opposite to bsr, which has a special
case when the operand is 0).
lzcnt is guarded by the ABM CPUID bit (Fn8000_0001:ECX_5).
Aurelien Jarno [Thu, 15 Oct 2009 21:08:46 +0000 (23:08 +0200)]
target-arm: fix sdiv helper
(INT32_MIN / -1) triggers an overflow, and the result depends on the
host architecture (INT32_MIN on arm, -1 on ppc, SIGFPE on x86). Use a
test to output the correct value.
Filip Navara [Thu, 15 Oct 2009 10:55:34 +0000 (12:55 +0200)]
target-arm: replace thumb usage of cpu_T registers by proper register allocations
The goal is eventually to get rid of all cpu_T register usage and to use
just short-lived tmp/tmp2 registers. This patch converts all the places where
cpu_T was used in the Thumb code and replaces it with explicit TCG register
allocation.
Filip Navara [Thu, 15 Oct 2009 10:45:57 +0000 (12:45 +0200)]
target-arm: convert NEON VZIP/VUZP/VTRN helper functions to pure TCG
The neon_trn_u8, neon_trn_u16, neon_unzip_u8, neon_zip_u8 and neon_zip_u16
helpers used fixed registers to return values. This patch replaces that with
TCG code, so T0/T1 is no longer directly used by the helper functions.
Bugs in the gen_neon_unzip register load code were also fixed.
Filip Navara [Thu, 15 Oct 2009 10:12:11 +0000 (12:12 +0200)]
target-arm: fix SRS/RFE instructions
The encoding of 'IA' and 'DB' conditions was swapped.
SRS instruction must store banked SPSR instead of CPSR at the specific address.
Missing 'return' statement at the end of RFE handling.
Fixed write-back code to reference correct registers.
Filip Navara [Thu, 15 Oct 2009 10:00:41 +0000 (12:00 +0200)]
target-arm: use tcg_global_mem_new_i32 to allocate registers
Currently each read/write of ARM register involves a LD/ST TCG operation. This
patch uses TCG memory-backed registers to represent the ARM register set. With
memory-backed registers the LD/ST operations are transparently generated by TCG
and host registers could be used to optimize the generated code.
Jan-Simon Möller [Sat, 17 Oct 2009 18:52:43 +0000 (21:52 +0300)]
Re: linux-user/syscall.c - don't add GUEST_BASE to NULL pointer
This patch fixes the mount call. GUEST_BASE shouldn't be added to a NULL pointer on arg5 . failing call:
mount("rootfs", "/", 0x47a78, MS_MGC_VAL|MS_REMOUNT, 0x10000) = -1 EFAULT (Bad address)
Izik Eidus [Thu, 8 Oct 2009 14:39:39 +0000 (16:39 +0200)]
ksm support
Call MADV_MERGEABLE on guest memory allocations. MADV_MERGABLE will be
available starting in Linux 2.6.32. This system call registers a region of
virtual address space with Linux as a candidate for transparent memory
sharing.
Kevin Wolf [Thu, 8 Oct 2009 13:02:08 +0000 (15:02 +0200)]
qcow2: Bring synchronous read/write back to life
When the synchronous read and write functions were dropped, they were replaced
by generic emulation functions. Unfortunately, these emulation functions don't
provide the same semantics as the original functions did.
The original bdrv_read would mean that we read some data synchronously and that
we won't be interrupted during this read. The latter assumption is no longer
true with the emulation function which needs to use qemu_aio_poll and therefore
allows the callback of any other concurrent AIO request to be run during the
read. Which in turn means that (meta)data read earlier could have changed and
be invalid now. qcow2 is not prepared to work in this way and it's just scary
how many places there are where other requests could run.
I'm not sure yet where exactly it breaks, but you'll see breakage with virtio
on qcow2 with a backing file. Providing synchronous functions again fixes the
problem for me.
Mark McLoughlin [Thu, 8 Oct 2009 18:58:31 +0000 (19:58 +0100)]
net: refactor packet queueing code
The packet queue code is fairly standalone, has some complex details and
easily reusable. It makes sense to split it out on its own. This patch
doesn't contain any functional changes.
Mark McLoughlin [Thu, 8 Oct 2009 18:58:30 +0000 (19:58 +0100)]
net: allow NICs to be connected to netdevs
Introduce a 'peer' member to VLANClientState as an alternative
to a vlan. The idea being that packets are transfered directly
from peer clients rather than going through a vlan.
Mark McLoughlin [Thu, 8 Oct 2009 18:58:27 +0000 (19:58 +0100)]
net: handle -netdevice options
Same as for -net except for:
- only tap, user, vde and socket types are supported
- the vlan parameter is not allowed
- the name parameter is not allowed but the id parameter is
required
Mark McLoughlin [Thu, 8 Oct 2009 18:58:24 +0000 (19:58 +0100)]
net: allow clients not associated with a vlan
Allow net clients to be created which are not connected to any vlan.
This is needed by Gerd in order to allow adding -device nic, where
the absence of a vlan parameter will not imply vlan=0. Also needed
to allow adding a -netdevice option which doesn't connect the backend
to a vlan.
Mark McLoughlin [Thu, 8 Oct 2009 18:58:22 +0000 (19:58 +0100)]
net: remove id field from NICInfo
Just use the name field instead since we now use the id paramater as
the name, if supplied. Only implication with this change is that if
id is not supplied, the value of the name paramater is used as an
id.
Stefan Weil [Fri, 9 Oct 2009 17:49:58 +0000 (19:49 +0200)]
eepro100: Remove unused device status entries
Once upon the time when QEMU hacking was fun
there was a brave knight who wanted to have
a driver for a special intel nic.
So he started by cloning ne2000.c which also
meant that the new born eepro100.c was
immediately three years old.
Other knights who also wanted to have fun and
take their part in the battle thought that it
would be a good idea to remove stupid code
which says "missing nic load, missing nic save".
They saved everything they saw, man and women,
ne2000 code and runtime address offsets, and
put all saved elements in a prison called
vm data.
When the first knight came back and noticed
the unhappy prisoners, he wanted to set them
free. But the keepers of the keys told him
that they would have to stay there forever
for compatibility reasons.
So our brave knight now takes a new effort
to save the souls of the poor prisoners by
removing their names.
Their bodies will have to rot in the dungeons
of compatibility forever, watched by the
keepers of the keys.
Artyom Tarasenko [Tue, 13 Oct 2009 21:46:24 +0000 (23:46 +0200)]
scsi disk block descriptor v2
The SCSI-2 documentation suggests, that although the block
descriptor is optional for an arbitrary SCSI-2 device (chapter 8.2.10,
http://ldkelley.com/SCSI2/SCSI2/SCSI2/SCSI2/SCSI2-08.html )
it is mandatory for a disk: chapters 9.1.2, 9.3.3
( http://ldkelley.com/SCSI2/SCSI2/SCSI2/SCSI2-09.html ) don't say
"optional" any more, just "The block descriptor in the MODE SENSE
data describes the block lengths that are used on the medium."
v2: limit the number of sectors reported in the block descriptor to 24 bits.
François Revol [Tue, 25 Aug 2009 09:14:10 +0000 (11:14 +0200)]
Fixed wacom emulation
- for absolute mode, scale coordinates to the real device maximum values,
since some drivers (on Haiku and Linux at least) need them as such,
and the HID descriptor is boggus on some models anyway,
- keep the coordinates even when no button is pressed, on real tablet
the pen is sensed on the surface even without direct contact,
and drivers expect this,
- map left button to pressure according to what the Haiku driver wants,
- map the right button to the pen button,
- map the middle button to the eraser,
- use asynchronous reporting as the hid code does, stops the Haiku driver
(and probably others) from spending 50% cpu polling for changes.
Gerd Hoffmann [Wed, 7 Oct 2009 11:37:06 +0000 (13:37 +0200)]
rom loader: fix sparc -kernel boot.
Changes:
(1) register pstrcpy_targphys() in rom list, it is used for kernel
command lines by a number of architectures.
(2) add rom_ptr() function to get a pointer for applying changes
to loaded images. Needed for example to tell the linux kernel
where it finds the initrd image by updating the header.
(3) make sparc use rom_ptr for initrd setup.
booting sparc-test works now, and 'info roms' shows this:
projects / qemu.git / log