Hervé Poussineau [Mon, 20 May 2013 10:33:08 +0000 (12:33 +0200)]
osdep: fix qemu_anon_ram_free trace (+ fix compilation on 32 bit hosts)
Commit e7a09b92b70786f9e8c5fbf787e0248c6ebbe707 added a trace at each
memory freeing, but unfortunately inverted size and pointer when printing
them. Fix trace.
This also led to a compilation error on 32 bit hosts:
In file included from include/trace.h:4:0,
from trace/generated-events.c:3:
./trace/generated-tracers.h: In function ‘trace_qemu_anon_ram_free’:
./trace/generated-tracers.h:64:9: error: format ‘%zu’ expects argument of type
‘size_t’, but argument 3 has type ‘void *’ [-Werror=format]
./trace/generated-tracers.h:64:9: error: format ‘%p’ expects argument of type
‘void *’, but argument 4 has type ‘size_t’ [-Werror=format]
Lei Li [Mon, 20 May 2013 06:51:03 +0000 (14:51 +0800)]
chardev: Make consistent with udp device for new qapi backend
When register and open a chardev udp, the backend name should be udp
not dgram, and we do not have backend dgram in the chardev list. This
patch makes the new qapi udp backend consistent with the original
udp device.
Petar Jovanovic [Sat, 18 May 2013 01:53:41 +0000 (03:53 +0200)]
target-mips: fix EXTPDP and setting up pos field in the DSPControl reg
This change makes sure that modifications of pos field in the DSPControl
register do not trash other bits in the register. This bug can be triggered
with the additional test case in mips32-dsp/extpdp.c in this commit.
In addition to this, this change corrects incorrect calculation of the mask
for EXTPDP.
Anthony Liguori [Fri, 17 May 2013 15:00:30 +0000 (10:00 -0500)]
Merge remote-tracking branch 'mdroth/qga-pull-2013-05-13' into staging
* mdroth/qga-pull-2013-05-13:
qga: unlink just created guest-file if fchmod() or fdopen() fails on it
qga: distinguish binary modes in "guest_file_open_modes" map
Reverting this patch fixes a divide-by-zero error in qemu that can be
fairly reliably triggered by doing block migration. In this case, the
configuration/error was:
This revert potentially re-introduces a bug that was present in 1.4,
but fixes a prevalent issue with block migration so we should revert
it for now and take an updated patch later.
Conflicts:
migration.c
* fixed up to remove logic introduced in 7161082c while leaving
changes in HEAD intact
pollfds_fill() and pollfds_poll() translate GPollFD to rfds/wfds/xfds
for sockets on win32. select(2) is the underlying system call which is
used to monitor sockets for activity.
Currently file descriptors that monitor G_IO_ERR will be included in
both rfds and wfds. As a result, select(2) will report writability on
file descriptors where we only really wanted to monitor readability
(with errors).
slirp_pollfds_poll() hit this issue: UDP sockets are blocking sockets so
we hang in sorecvfrom() when G_IO_ERR is set due to the socket being
writable (we only wanted to check for readability).
Anthony Liguori [Wed, 15 May 2013 19:57:30 +0000 (14:57 -0500)]
Merge remote-tracking branch 'luiz/queue/qmp' into staging
# By Michael Roth (1) and Zhangleiqiang (1)
# Via Luiz Capitulino
* luiz/queue/qmp:
qapi: fix leak in unit tests
qmp: fix handling of cmd with Equals in qmp-shell
Paolo Bonzini [Wed, 15 May 2013 15:46:11 +0000 (17:46 +0200)]
qemu-common: Resolve vector build breakes for AltiVec
On Mac OS X ppc, altivec.h defines "vector", leading to build breakage
when used as variable name, e.g. in tracing code.
Fix this by undefining identifiers after altivec.h inclusion.
Peter Maydell [Tue, 14 May 2013 15:33:36 +0000 (16:33 +0100)]
hw/pci-host/versatile.c: Provide property for forcing broken IRQ mapping
Although we try our best to automatically detect broken versions
of Linux which assume the old broken IRQ mapping we used to implement
for our model of the Versatile PCI controller, it turns out that
some particularly new kernels manage to outwit the autodetection.
We therefore provide a property for enabling the old broken IRQ
mapping, so that if users happen to have such a kernel they can
work around its deficiencies with the command line option:
-global versatile_pci.broken-irq-mapping=1
Peter Maydell [Tue, 14 May 2013 15:33:35 +0000 (16:33 +0100)]
hw/pci-host/versatile.c: Update autodetect to detect newer kernels
Newer versatilepb kernels still don't get the IRQ mapping right
for the PCI controller, but they get it differently wrong (they add
a fixed +64 offset to everything they write to PCI_INTERRUPT_LINE).
Update the autodetection to handle these too, and include a more
detailed comment on the various different behaviours that might
be present.
Peter Maydell [Tue, 14 May 2013 15:33:34 +0000 (16:33 +0100)]
Revert "versatile_pci: Put the host bridge PCI device at slot 29"
This reverts commit 5f37ef92b7690423ac6311d3c597e182fc5f8fe6.
It turns out that some kernels incorrectly depend on the
old QEMU behaviour of not putting the host PCI bridge device
where the hardware puts it, because they use a swizzling IRQ
mapping which is incorrect but happens to match up with old
broken QEMU when the slot number mod 4 is zero. Since we
start PCI devices at 11, if we put the host bridge at 29
then the first real PCI device goes at 11 and doesn't work.
Not putting the host bridge at 29 means it defaults to 11,
so the first real PCI device is at 12 and works.
Since continuing with the old behaviour doesn't cause problems
for kernels which do work with hardware, the simplest fix for
this is to revert the change.
Peter Maydell [Tue, 14 May 2013 20:36:39 +0000 (21:36 +0100)]
configure: Detect uuid on MacOSX (fixes compile failure)
Commit 7791dba3ec broke compilation on MacOSX, because it introduced
a new include of util.h. On MacOSX this includes pwd.h which in turn
includes the system uuid/uuid.h, which causes a compile failure if
QEMU was configured without CONFIG_UUID due to a conflict between
the system header and our fallback versions:
block/vdi.c:124:20: error: static declaration of 'uuid_generate'
follows non-static declaration
static inline void uuid_generate(uuid_t out)
^
/usr/include/uuid/uuid.h:63:6: note: previous declaration is here
void uuid_generate(uuid_t out);
^
Fix this breakage by improving configure's check for uuid to work on
MacOSX (where there is no need to link in a separate libuuid).
Note that if the user explicitly runs configure with '--disable-uuid'
on MacOSX then QEMU will fail to compile.
Michael Roth [Fri, 10 May 2013 02:20:57 +0000 (21:20 -0500)]
qapi: fix leak in unit tests
qmp_output_get_qobject() increments the qobject's reference count. Since
we currently pass this straight into qobject_to_json() so we can feed
the data into a QMP input visitor, we never actually free the underlying
qobject when qmp_output_visitor_cleanup() is called. This causes leaks
on all of the QMP serialization tests.
Fix this by holding a pointer to the qobject and decref'ing it before
returning from qmp_deserialize().
Zhangleiqiang [Mon, 6 May 2013 08:31:23 +0000 (08:31 +0000)]
qmp: fix handling of cmd with Equals in qmp-shell
qmp: fix handling of cmd with equal mark in qmp-shell
qmp-shell splits the argument and value of input command
by equal mark("="). But there are commands whose values
include equal mark themselves, and the json built by
qmp-shell will not correct. For example, when using NBD as
the target of block-backup command, the input
"block-backup target=nbd+unix:///drive0?socket=/tmp/nbd.sock"
will fail, because the json built will be as follows:
With this patch applied the segfault will be avoided, however the case
will still fail, though gracefully:
$ qemu-img create -f qcow2 /tmp/huge.qcow2 $((1024*1024))T
Formatting 'huge.qcow2', fmt=qcow2 size=1152921504606846976 encryption=off cluster_size=65536 lazy_refcounts=off
qemu-img: The image size is too large for file format 'qcow2'
Note that even long before these overflow checks kick in, you get
insanely high memory usage (up to INT_MAX * sizeof(uint64_t) = 16 GB for
the L1 table), so with somewhat smaller image sizes you'll probably see
qemu aborting for a failed g_malloc().
If you need huge image sizes, you should increase the cluster size to
the maximum of 2 MB in order to get higher limits.
Paolo Bonzini [Mon, 13 May 2013 14:19:56 +0000 (16:19 +0200)]
osdep: introduce qemu_anon_ram_free to free qemu_anon_ram_alloc-ed memory
We switched from qemu_memalign to mmap() but then we don't modify
qemu_vfree() to do a munmap() over free(). Which we cannot do
because qemu_vfree() frees memory allocated by qemu_{mem,block}align.
Introduce a new function that does the munmap(), luckily the size is
available in the RAMBlock.
Paolo Bonzini [Tue, 14 May 2013 12:12:25 +0000 (14:12 +0200)]
spitz: fix compilation failure due to pty.h namespace pollution
pty.h is polluting the global namespace with a CTRL macro. spitz
thus fails compilation with the patch at
http://article.gmane.org/gmane.comp.emulators.qemu/211337 and
this patch fixes it.
Anthony Liguori [Mon, 13 May 2013 20:22:24 +0000 (15:22 -0500)]
qom: aggressively optimize qom casting
This patch adds a small typename cache to ObjectClass. This allows
caching positive casts within each ObjectClass. Benchmarking a
PPC workload provided by Aurelien, this patch eliminates every
single g_hash_table_lookup() happening during the benchmark (which
was about 2 million per-second).
With this patch applied, I get exactly the same performance (within
the margin of error) as with --disable-qom-cast-debug.
N.B. it's safe to cache typenames only from the _assert() macros
because they are always called with string literals.
Igor Mitsyanko [Fri, 10 May 2013 14:59:45 +0000 (18:59 +0400)]
ui/gtk.c: do not use gdk_display_warp_pointer when GTK ver >3.0
Commit 9697f5d2d38e5dd1e64e8e0d64436e6d44e7b1fe "gtk: custom cursor support"
introduced unconditional usage of gdk_display_warp_pointer(). This function
is marked as deprecated since GTK-3.0, and triggers warning (error with -Werror)
during compilation.
Conditionally change gdk_display_warp_pointer() method usage to gdk_device_warp
usage, as suggested by compiler.
Paolo Bonzini [Fri, 10 May 2013 12:38:03 +0000 (14:38 +0200)]
Revert "pc: Kill the "use flash device for BIOS unless KVM" misfeature"
This reverts commit 9953f8822cc316eec9962f0a2858c3439a80adec.
While Markus's analysis is entirely correct, there are 1.6 patches
that fix the bug for real and without requiring machine type hacks.
Let's think of the children who will have to read this code, and
avoid a complicated mess of semantics that differ between <1.5,
1.5, and >1.5.
Paolo Bonzini [Fri, 10 May 2013 12:16:40 +0000 (14:16 +0200)]
qom: allow turning cast debugging off
Cast debugging can have a substantial cost (20% or more). Instead of adding
special-cased "fast casts" in the hot paths, we can just disable it in
releases. The tracing facilities we just added make it easier to analyze
those problems that cast debugging would reveal.
Paolo Bonzini [Fri, 10 May 2013 12:16:37 +0000 (14:16 +0200)]
qom: add a fast path to object_class_dynamic_cast
For leaf classes, in many cases the callbacks will simply downcast
the object back to the original class. Add this fast path to
object_class_dynamic_cast, object_dynamic_cast will inherit it.
Laszlo Ersek [Wed, 8 May 2013 15:31:35 +0000 (17:31 +0200)]
qga: distinguish binary modes in "guest_file_open_modes" map
In Windows guests this may make a difference.
Since the original patch (commit c689b4f1) sought to be pedantic and to
consider theoretical corner cases of portability, we should fix it up
where it failed to come through in that pursuit.
Anthony Liguori [Mon, 13 May 2013 14:44:25 +0000 (09:44 -0500)]
Merge remote-tracking branch 'mjt/trivial-patches-next' into staging
# By Dong Xu Wang (2) and others
# Via Michael Tokarev
* mjt/trivial-patches-next:
osdep.h: include sys/types.h for ssize_t definition
remove double semicolons
clean unnecessary code: don't check g_strdup arg for NULL
docs: mention AddressSpaces in docs/memory.txt
audio: update documentation after removing --audio-card-list option
m25p80.c: Sync Flash chip list with Linux
bsd-user: OS-agnostic 64-bit SYSCTL types
Igor Mitsyanko [Fri, 10 May 2013 19:58:21 +0000 (23:58 +0400)]
osdep.h: include sys/types.h for ssize_t definition
sys/types.h is taken out from "ifdef __OpenBSD__" guard. It should be
safe for other systems, according to following survey:
http://hacks.owlfolio.org/header-survey/
This fixes build for CONFIG_IOVEC-less systems (mingw).
Ed Maste [Wed, 1 May 2013 14:40:53 +0000 (10:40 -0400)]
m25p80.c: Sync Flash chip list with Linux
Add new devices for various manufacturers, and re-sort Spansion list to
match the order in Linux, which requires chips with a non-zero extended ID
to come first.
With this commit the outstanding differences to Linux rev 55bf75b are:
- Erase size flag differences in s25sl032p, s25sl064p, s25fl016k, s25fl064k
(These devices have only some blocks that support small erase sizes.)
- Linux lacks n25q128
- Devices without a Jedec ID have been excluded
Ed Maste [Tue, 30 Apr 2013 13:29:32 +0000 (09:29 -0400)]
bsd-user: OS-agnostic 64-bit SYSCTL types
Use existence of type as #ifdef condition rather than FreeBSD-specific
version check, as suggested by Patrick Welche.
Also handle the signed (CTLTYPE_S64) case identically to the unsigned
(CTLTYPE_U64) case, per later patches in the FreeBSD ports tree
(emulators/qemu-devel/files/patch-z-arm-bsd-user-001).
Aurelien Jarno [Wed, 8 May 2013 20:36:39 +0000 (22:36 +0200)]
tcg/optimize: fix setcond2 optimization
When setcond2 is rewritten into setcond, the state of the destination
temp should be reset, so that a copy of the previous value is not
used instead of the result.
Hans de Goede [Tue, 23 Apr 2013 08:18:16 +0000 (10:18 +0200)]
qxl: Call spice_qxl_driver_unload from qxl_enter_vga_mode
With a SPICE_DISPLAY_CAP_MONITORS_CONFIG capable client, the client needs to
know what part of the primary to use for each monitor. If the guest driver
does not support this, the server sends messages to the client for a
single monitor spanning the entire primary.
As soon as the guest calls spice_qxl_monitors_config_async once, the server
sets the red_worker driver_has_monitors_config flag and stops doing this.
This is a problem when the driver gets unloaded, for example after a reboot
or when switching to a text vc with usermode mode-setting under Linux.
To reproduce this start a multi-mon capable Linux guest which uses
usermode mode-setting and then once X has started switch to a text vc. Note
how the client window does not only not resize, if you try to resize it
manually you always keep blackborders since the aspect is wrong.
This patch calls a new spice-server method called spice_qxl_driver_unload
which clears the driver_has_monitors_config flag inside the server, thereby
fixing this.
Anthony Liguori [Wed, 8 May 2013 20:54:36 +0000 (15:54 -0500)]
Merge remote-tracking branch 'kwolf/for-anthony' into staging
# By Kevin Wolf (7) and Fam Zheng (3)
# Via Kevin Wolf
* kwolf/for-anthony:
qemu-iotests: fix 017 018 for vmdk
qemu-iotests: exclude vmdk and qcow from 043
qemu-iotests: exclude vmdk for test 042
qtest/ide-test: Test short and long PRDTs
qtest/ide-test: Add simple DMA read/write test case
qtest: Add IDE test case
libqos/pci: Enable bus mastering
ide: Reset BMIDEA bit when the bus master is stopped
de_DE.po: Add missing leading spaces
ahci: Don't allow creating slave drives
Jason Wang [Tue, 7 May 2013 05:42:49 +0000 (13:42 +0800)]
virtio: properly validate address before accessing config
There are several several issues in the current checking:
- The check was based on the minus of unsigned values which can overflow
- It was done after .{set|get}_config() which can lead crash when config_len
is zero since vdev->config is NULL
Fix this by:
- Validate the address in virtio_pci_config_{read|write}() before
.{set|get}_config
- Use addition instead minus to do the validation
mask notifiers are never called without msix,
so devices with backend masking like vhost don't work.
Call mask notifiers explicitly at
startup/cleanup to make it work.
Alexander Graf [Mon, 6 May 2013 17:49:43 +0000 (19:49 +0200)]
PPC: Fix rldcl
The implementation for rldcl tried to always fetch its
parameters from the opcode, even though the opcode was
already passed in in decoded and different forms.
Alexander Graf [Wed, 8 May 2013 13:19:14 +0000 (15:19 +0200)]
PPC: Depend behavior of cmp instructions only on instruction encoding
When running an L=1 cmp instruction on a 64bit PPC CPU with SF off, it
still behaves identical to what it does when SF is on. Remove the implicit
difference in the code.
Also, on most 32bit CPUs we should always treat the compare as 32bit
compare, as the CPU will ignore the L bit. This is not true for e500mc,
but that's up for a different patch.
Petar Jovanovic [Wed, 8 May 2013 11:17:40 +0000 (13:17 +0200)]
target-mips: fix incorrect behaviour for INSV
Corner case for INSV instruction when size=32 has not been correctly
implemented. The mask for size should be one bit wider, and preparing the
filter variable should be aware of this case too.
The test for INSV has been extended to include the case that triggers the
bug.
Fam Zheng [Tue, 7 May 2013 11:13:05 +0000 (19:13 +0800)]
qemu-iotests: fix 017 018 for vmdk
017 and 018 use /bin/mv to move base img from t.IMGFMG to t.IMGFMT.base
after filling data, this is not enough for vmdk, when t.IMGFMT is only a
description text file who points to t-{flat,s001,f001,...}.IMGFMT as
data extent, so testing such subformats alway fails on them.
This patch use the trick of temprorily changing TEST_IMG to avoid using
/bin/mv.
Kevin Wolf [Wed, 13 Mar 2013 12:30:24 +0000 (13:30 +0100)]
qtest/ide-test: Test short and long PRDTs
This tests the behaviour of the DMA engine when the given PRDT contains
physical region descriptors for either more or less bytes than the
IDE request is for.
Kevin Wolf [Wed, 8 May 2013 09:18:41 +0000 (11:18 +0200)]
qtest: Add IDE test case
This adds a simple IDE test case and starts by verifying that IDENTIFY
can be successfully used and return the correct serial number, version
and the WCE flag is set for cache=writeback.
Kevin Wolf [Wed, 13 Mar 2013 15:53:06 +0000 (16:53 +0100)]
ide: Reset BMIDEA bit when the bus master is stopped
The device is supposed to reset the Bus Master IDE Active bit in the
status register when 0 is written to the Start/Stop Bus Master bit in
the command register.
In the common cases this happens automatically because bdrv_drain_all()
flushes the requests, but with a large PRDT it could remain set.
Kevin Wolf [Mon, 6 May 2013 13:58:04 +0000 (15:58 +0200)]
ahci: Don't allow creating slave drives
An IDE bus provided by AHCI can only take a single IDE drive. If you add
a drive as slave, qemu used to accept the command line but the device
wouldn't be actually usable. Catch the situation instead and error out.
Anthony Liguori [Tue, 7 May 2013 11:48:31 +0000 (06:48 -0500)]
Merge remote-tracking branch 'kraxel/usb.82' into staging
# By Gerd Hoffmann (3) and Hans de Goede (1)
# Via Gerd Hoffmann
* kraxel/usb.82:
xhci: handle USB_RET_BABBLE
uhci: Use an intermediate buffer for usb packet data
usb-host: add usb_host_full_speed_compat
usb-host: live migration support for the libusb version
In addition, at least all files created with the "guest-file-open" QMP
command, and all files created with shell output redirection (or
otherwise) by utilities invoked by the fsfreeze hook script are affected.
For now mask all file mode bits for "group" and "others" in
become_daemon().
Temporarily, for compatibility reasons, stick with the 0666 file-mode in
case of files newly created by the "guest-file-open" QMP call. Do so
without changing the umask temporarily.
Jason Wang [Fri, 26 Apr 2013 10:27:40 +0000 (18:27 +0800)]
virtio-net: properly check the vhost status during status set
Commit 32993698 (vhost: disable on tap link down) tries to disable the vhost
also when the peer's link is down. But the check was not done properly, the
vhost were only started when:
1) peer's link is not down
2) virtio-net has already been started.
Since == have a higher precedence than &&, place a brace to make sure both the
conditions were met then does the check. This fixes the crash when doing a savem
after set the link off which let qemu crash and complains:
Anthony Liguori [Tue, 7 May 2013 11:32:00 +0000 (06:32 -0500)]
Merge remote-tracking branch 'afaerber/qom-cpu' into staging
# By Eduardo Habkost (6) and others
# Via Andreas Färber
* afaerber/qom-cpu:
target-i386: n270 can MOVBE
target-i386: Introduce generic CPUID feature compat function
target-i386: Change CPUID model of 486 to 8
target-i386: Emulate X86CPU subclasses for global properties
qdev: Introduce qdev_prop_set_globals_for_type()
qdev: Let qdev_prop_parse() pass through Error
target-i386: Add "filtered-features" property to X86CPU
target-i386: Introduce X86CPU::filtered_features field
target-i386: Add "feature-words" property to X86CPU
target-i386: Use FeatureWord loop on filter_features_for_kvm()
target-i386: Add ECX information to FeatureWordInfo
Hans de Goede [Mon, 6 May 2013 08:48:57 +0000 (10:48 +0200)]
uhci: Use an intermediate buffer for usb packet data
Due to various unfortunate reasons we cannot reliable detect a guest
cancelling a packet as soon as it happens, instead we detect cancels
with some delay.
When packets are handled async, and we directly pass the guest memory for
the packet to the usb-device as iovec, this means that the usb-device can
write to guest-memory which the guest has already re-used for other purposes
-> not good!
This patch fixes this by adding an intermediate buffer and writing back not
only the result, but also the data, of async completed packets when scanning
the schedule.
Gerd Hoffmann [Mon, 6 May 2013 11:12:16 +0000 (13:12 +0200)]
usb-host: add usb_host_full_speed_compat
Alloes to pass through usb2 devices on usb1 host controllers if possible.
Brings the libusb implementation to feature-parity with the linux usbfs
code, so the usb-host implementation in 1.5 (libusb) doesn't regress
compared to 1.4 (usbfs).
The Atom core (cpu name "n270" in QEMU speak) supports MOVBE. This is
needed when booting 3.8 and later linux kernels built with the MATOM
target because we require MOVBE in order to boot properly now.
Signed-off-by: Borislav Petkov <[email protected]>
[ehabkost: added compat code to disable MOVBE on pc-*-1.4 and older] Signed-off-by: Eduardo Habkost <[email protected]> Signed-off-by: Andreas Färber <[email protected]>
Aurelien Jarno [Mon, 6 May 2013 17:56:27 +0000 (19:56 +0200)]
Merge branch 's390-for-upstream' of git://github.com/agraf/qemu
* 's390-for-upstream' of git://github.com/agraf/qemu:
s390: update s390-ccw.img
S390: BIOS boot from given device
S390: Add virtio-blk boot
S390: Merging s390_ipl_cpu and s390_ipl_reset
S390: BIOS create link to src folder for .img file
S390: BIOS check for file
Andreas Färber [Wed, 1 May 2013 15:05:47 +0000 (17:05 +0200)]
target-i386: Emulate X86CPU subclasses for global properties
After initializing the object from its x86_def_t and before setting any
additional -cpu arguments, set any global properties for the designated
subclass <name>-{i386,x86_64}-cpu.
Eduardo Habkost [Mon, 6 May 2013 16:20:09 +0000 (13:20 -0300)]
target-i386: Add "filtered-features" property to X86CPU
This property will contain all the features that were removed from the
CPU because they are not supported by the host.
This way, libvirt or other management tools can emulate the
check/enforce behavior by checking if filtered-properties is all zeroes,
before starting the guest.
Eduardo Habkost [Mon, 6 May 2013 16:20:07 +0000 (13:20 -0300)]
target-i386: Add "feature-words" property to X86CPU
This property will be useful for libvirt, as libvirt already has logic
based on low-level feature bits (not feature names), so it will be
really easy to convert the current libvirt logic to something using the
"feature-words" property.
The property will have two main use cases:
- Checking host capabilities, by checking the features of the "host"
CPU model
- Checking which features are enabled on each CPU model
target-i386: Use FeatureWord loop on filter_features_for_kvm()
Instead of open-coding the filtering code for each feature word, change
the existing code to use the feature_word_info array, that has exactly
the same CPUID eax/ecx/register values for each feature word.
projects / qemu.git / log