Setting the client architecture DHCP option to 0x001f (s390 Basic) [1]
allows the DHCP server to return a s390-specific bootfile if wanted.
DHCP servers not configured for the option (or not yet recognizing the
option value) will continue to work as they have done before.
Halil Pasic [Fri, 8 Sep 2017 15:24:44 +0000 (17:24 +0200)]
s390x/css: fix NULL handling for CCW addresses
Back then in the time of df1fe5bb49 ("s390: Virtual channel subsystem
support.", 2013-01-24) -EIO used to map to a channel-program check (via
the default label of the switch statement). Then 2dc95b4cac
("s390x/3270: 3270 data stream handling", 2016-04-01) came along
and that changed dramatically.
Let us roll back this undesired side effect, and go back to
channel-program check.
Halil Pasic [Fri, 8 Sep 2017 15:24:43 +0000 (17:24 +0200)]
s390x/css: drop data-check in interpretation
The architecture says that channel-data check is indicating that
an uncorrected storage (memory) error has been detected in regard
to the data residing in main storage (memory) that is currently
used for an I/O operation. The described detection is done using
the CBC technology.
The ccw interpretation code is however generating a channel-data check
effectively when the (device specific) ccw_cb returns -EFAULT. In case
of virtio-ccw devices this happens when mapping memory fails, or when a
NULL pointer is encountered. So this behavior is not architecture
conform.
Furthermore the best fit for these situations (null pointer, mapping a
piece of guest memory fails) from architectural perspective the condition
described as the channel subsystem refers to a location that is not
available, which when encountered shall result in a channel-program
check.
To fix this, all we have to do is to get rid of the switch case matching
-EFAULT: the default is generating a channel-program check.
Thomas Huth [Wed, 30 Aug 2017 13:39:03 +0000 (15:39 +0200)]
hw/misc/ivshmem: Fix ivshmem_recv_msg() to also work on big endian systems
The "slow" ivshmem-tests currently fail when they are running on a
big endian host:
$ uname -m
ppc64
$ V=1 QTEST_QEMU_BINARY=x86_64-softmmu/qemu-system-x86_64 tests/ivshmem-test -m slow
/x86_64/ivshmem/single: OK
/x86_64/ivshmem/hotplug: OK
/x86_64/ivshmem/memdev: OK
/x86_64/ivshmem/pair: OK
/x86_64/ivshmem/server-msi: qemu-system-x86_64:
-device ivshmem-doorbell,chardev=chr0,vectors=2: server sent invalid ID message
Broken pipe
The problem is that the server side code in ivshmem_server_send_one_msg()
correctly translates all messages IDs into little endian 64-bit values,
but the client side code in the ivshmem_recv_msg() function does not swap
the byte order back. Fix it by passing the value through le64_to_cpu().
Yi Min Zhao [Tue, 5 Sep 2017 10:13:00 +0000 (12:13 +0200)]
s390x/pci: add iommu replay callback
Let's introduce iommu replay callback for s390 pci iommu memory region.
Currently we don't need any dma mapping replay. So let it return
directly. This implementation will avoid meaningless loops calling
translation callback.
Yi Min Zhao [Tue, 5 Sep 2017 10:12:59 +0000 (12:12 +0200)]
s390x/pci: fixup ind_offset of msix routing entry
The guest uses the mpcifc instruction to register the aibvo of a zpci
device, which is the starting offset of indicators in the indicator
area and thus remains constant. Each msix vector is an offset from the
aibvo. When we map a msix route to an adapter route, we should not
modify the starting offset, but instead add the vector to the starting
offset to get the absolute offset in the specific route.
Yi Min Zhao [Tue, 5 Sep 2017 10:12:58 +0000 (12:12 +0200)]
s390x/pci: remove idx from msix msg data
PCIDevice pointer has been a parameter of kvm_arch_fixup_msi_route().
So we don't need to store zpci idx in msix message data to find out the
specific zpci device. Instead, we could use pci device id to find its
corresponding zpci device.
Thomas Huth [Thu, 31 Aug 2017 14:40:08 +0000 (16:40 +0200)]
tests: Enable the drive_del test also on s390x
We can use the drive_del test on s390x, too, to check that adding and
deleting also works fine with the virtio-ccw bus. But we have to make
sure that we use the devices with the "-ccw" suffix instead of the
"-pci" suffix for the virtio-ccw transport on s390x. Introduce a helper
function called qvirtio_get_dev_type() that returns the correct string
for the current architecture.
Halil Pasic [Thu, 31 Aug 2017 12:18:28 +0000 (14:18 +0200)]
s390x/css: fix cc handling for XSCH
The function ioinst_handle_xsch is presenting cc 2 when it's supposed to
present cc 1 and the other way around, because css_do_xsch has the error
codes mixed up. Because cc 1 has precedence over cc 2 we also have to
swap the two checks.
* remotes/bonzini/tags/for-upstream: (51 commits)
docker: fix creation of archives
default-configs: Replace $(and ...) with $(call land, ...)
osdep.h: Prohibit disabling assert() in supported builds
checkpatch: add hwaddr to @typeList
accel/hax: move hax-stub.c to accel/stubs/
target/i386: fix "info mem" for LA57 mode
scripts: let checkpatch.pl process an entire GIT branch
update-linux-headers: prepare for hyperv.h removal
hyperv: add header with protocol definitions
i386/cpu/hyperv: support over 64 vcpus for windows guests
Convert remaining single line fprintf() to warn_report()
Makefile: Remove libqemustub.a
ptimer-test: do not link to libqemustub.a/libqemuutil.a
target/mips: Convert VM clock update prints to warn_report
General warn report fixups
Convert multi-line fprintf() to warn_report()
Convert single line fprintf(.../n) to warn_report()
Convert remaining error_report() to warn_report()
hw/i386: Improve some of the warning messages
test-qga: add missing qemu-ga tool dependency
...
Thomas Huth [Mon, 18 Sep 2017 18:32:18 +0000 (20:32 +0200)]
default-configs: Replace $(and ...) with $(call land, ...)
Using $(and ...) is dangerous here: It only works as long as the first
argument is set to 'y' or completely unset. It does not work if the
first argument is set to 'n' for example. Let's use the "land" make
function instead which has been written explicitely for this purpose.
Eric Blake [Mon, 11 Sep 2017 21:13:20 +0000 (16:13 -0500)]
osdep.h: Prohibit disabling assert() in supported builds
We already have several files that knowingly require assert()
to work, sometimes because refactoring the code for proper
error handling has not been tackled yet; there are probably
other files that have a similar situation but with no comments
documenting the same. In fact, we have places in migration
that handle untrusted input with assertions, where disabling
the assertions risks a worse security hole than the current
behavior of losing the guest to SIGABRT when migration fails
because of the assertion. Promote our current per-file
safety-valve to instead be project-wide, and expand it to also
cover glib's g_assert().
Note that we do NOT want to encourage 'assert(side-effects);'
(that is a bad practice that prevents copy-and-paste of code to
other projects that CAN disable assertions; plus it costs
unnecessary reviewer mental cycles to remember whether a project
special-cases the crippling of asserts); and we would LIKE to
fix migration to not rely on asserts (but that takes a big code
audit). But in the meantime, we DO want to send a message
that anyone that disables assertions has to tweak code in order
to compile, making it obvious that they are taking on additional
risk that we are not going to support. At the same time, leave
comments mentioning NDEBUG in files that we know still need to
be scrubbed, so there is at least something to grep for.
It would be possible to come up with some other mechanism for
doing runtime checking by default, but which does not abort
the program on failure, while leaving side effects in place
(unlike how crippling assert() avoids even the side effects),
perhaps under the name q_verify(); but it was not deemed worth
the effort (developers should not have to learn a replacement
when the standard C macro works just fine, and it would be a lot
of churn for little gain). The patch specifically uses #error
rather than #warn so that a user is forced to tweak the header
to acknowledge the issue, even when not using a -Werror
compilation.
Greg Kurz [Thu, 14 Sep 2017 09:12:07 +0000 (11:12 +0200)]
checkpatch: add hwaddr to @typeList
The script doesn't know about all possible types and learn them as
it parses the code. If it reaches a line with a type cast but the
type isn't known yet, it is misinterpreted as an identifier.
For example the following line:
foo = (hwaddr) -1;
results in the following false-positive to be reported:
ERROR: spaces required around that '-' (ctx:VxV)
Let's add this standard QEMU type to the list of pre-known types.
scripts: let checkpatch.pl process an entire GIT branch
Currently before submitting a series, devs should run checkpatch.pl
across each patch to be submitted. This can be automated using a
command such as:
git rebase -i master -x 'git show | ./scripts/checkpatch.pl -'
This is rather long winded to type, so this patch introduces a way
to tell checkpatch.pl to validate a series of GIT revisions.
There are now three modes it can operate in 1) check a patch 2) check a source
file, or 3) check a git branch.
If no flags are given, the mode is determined by checking the args passed to
the command. If the args contain a literal ".." it is treated as a GIT revision
list. If the args end in ".patch" or equal "-" it is treated as a patch file.
Otherwise it is treated as a source file.
This automatic guessing can be overridden using --[no-]patch --[no-]file or
--[no-]branch
Roman Kagan [Thu, 13 Jul 2017 20:15:21 +0000 (23:15 +0300)]
hyperv: add header with protocol definitions
The definitions for Hyper-V emulation are currently taken from a header
imported from the Linux kernel.
However, as these describe a third-party protocol rather than a kernel
API, it probably wasn't a good idea to publish it in the kernel uapi.
This patch introduces a header that provides all the necessary
definitions, superseding the one coming from the kernel.
The new header supports (temporary) coexistence with the kernel one.
The constants explicitly named in the Hyper-V specification (e.g. msr
numbers) are defined in a non-conflicting way. Other constants and
types have got new names.
While at this, the protocol data structures are defined in a more
conventional way, without bitfields, enums, and excessive unions.
The code using this stuff is adjusted, too; it can now be built both
with and without the kernel header in the tree.
i386/cpu/hyperv: support over 64 vcpus for windows guests
Starting with Windows Server 2012 and Windows 8, if
CPUID.40000005.EAX contains a value of -1, Windows assumes specific
limit to the number of VPs. In this case, Windows Server 2012
guest VMs may use more than 64 VPs, up to the maximum supported
number of processors applicable to the specific Windows
version being used.
Alistair Francis [Fri, 18 Aug 2017 18:47:35 +0000 (11:47 -0700)]
Convert remaining single line fprintf() to warn_report()
Convert any remaining uses of fprintf(stderr, "warning:"...
to use warn_report() instead. This helps standardise on a single
method of printing warnings to the user.
All of the warnings were changed using this command:
find ./* -type f -exec sed -i 's|fprintf(.*".*warning[,:] |warn_report("|Ig' {} +
The #include lines and chagnes to the test Makefile were manually
updated to allow the code to compile.
Paolo Bonzini [Tue, 19 Sep 2017 14:20:31 +0000 (16:20 +0200)]
Makefile: Remove libqemustub.a
Using two libraries (libqemuutil.a and libqemustub.a) would sometimes
result in circular dependencies. To avoid these issues let's just
combine both into a single library that functions as both.
Alistair Francis [Mon, 11 Sep 2017 19:52:53 +0000 (12:52 -0700)]
Convert multi-line fprintf() to warn_report()
Convert all the multi-line uses of fprintf(stderr, "warning:"..."\n"...
to use warn_report() instead. This helps standardise on a single
method of printing warnings to the user.
All of the warnings were changed using these commands:
find ./* -type f -exec sed -i \
'N; {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N; {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N;N; {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N;N;N {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N;N;N;N {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N;N;N;N;N {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N;N;N;N;N;N; {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
{} +
Indentation fixed up manually afterwards.
Some of the lines were manually edited to reduce the line length to below
80 charecters. Some of the lines with newlines in the middle of the
string were also manually edit to avoid checkpatch errrors.
The #include lines were manually updated to allow the code to compile.
Several of the warning messages can be improved after this patch, to
keep this patch mechanical this has been moved into a later patch.
Alistair Francis [Mon, 11 Sep 2017 19:52:50 +0000 (12:52 -0700)]
Convert single line fprintf(.../n) to warn_report()
Convert all the single line uses of fprintf(stderr, "warning:"..."\n"...
to use warn_report() instead. This helps standardise on a single
method of printing warnings to the user.
All of the warnings were changed using this command:
find ./* -type f -exec sed -i \
's|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig' \
{} +
Some of the lines were manually edited to reduce the line length to below
80 charecters.
The #include lines were manually updated to allow the code to compile.
Alistair Francis [Mon, 11 Sep 2017 19:52:46 +0000 (12:52 -0700)]
Convert remaining error_report() to warn_report()
In a previous patch (3dc6f8693694a649a9c83f1e2746565b47683923) we
converted uses of error_report("warning:"... to use warn_report()
instead. This was to help standardise on a single method of printing
warnings to the user.
There appears to have been some cases that slipped through in patch sets
applied around the same time, this patch catches the few remaining
cases.
All of the warnings were changed using this command:
find ./* -type f -exec sed -i \
's|error_report(".*warning[,:] |warn_report("|Ig' {} +
kvm: kvm_log_sync() is only called with known memory sections
Flatview will make sure that we can only end up in this function with
memory sections that correspond to exactly one slot. So we don't
have to iterate multiple times. There won't be overlapping slots but
only matching slots.
Properly align the section and look up the corresponding slot. This
heavily simplifies this function.
We can now get rid of kvm_lookup_overlapping_slot().
kvm: we never have overlapping slots in kvm_set_phys_mem()
The way flatview handles memory sections, we will never have overlapping
memory sections in kvm.
address_space_update_topology_pass() will make sure that we will only
get called for
a) an existing memory section for which we only update parameters
(log_start, log_stop).
b) an existing memory section we want to delete (region_del)
c) a brand new memory section we want to add (region_add)
We cannot have overlapping memory sections in kvm as we will first remove
the overlapping sections and then add the ones without conflicts.
Therefore we can remove the complexity for handling prefix and suffix
slots.
We already require DESTROY_MEMORY_REGION_WORKS, JOIN_MEMORY_REGIONS_WORKS
was added just half a year later.
In addition, with flatview overlapping memory regions are first
removed before adding the changed one. So we can't really detect joining
memory regions this way.
While loading kernel via multiboot-v1 image, (flags & 0x00010000)
indicates that multiboot header contains valid addresses to load
the kernel image. These addresses are used to compute kernel
size and kernel text offset in the OS image. Validate these
address values to avoid an OOB access issue.
Ladi Prosek [Mon, 7 Aug 2017 08:57:03 +0000 (10:57 +0200)]
i386/kvm: advertise Hyper-V frequency MSRs
As of kernel commit eb82feea59d6 ("KVM: hyperv: support HV_X64_MSR_TSC_FREQUENCY
and HV_X64_MSR_APIC_FREQUENCY"), KVM supports two new MSRs which are required
for nested Hyper-V to read timestamps with RDTSC + TSC page.
This commit makes QEMU advertise the MSRs with CPUID.40000003H:EAX[11] and
CPUID.40000003H:EDX[8] as specified in the Hyper-V TLFS and experimentally
verified on a Hyper-V host. The feature is enabled with the existing hv-time CPU
flag, and only if the TSC frequency is stable across migrations and known.
Ladi Prosek [Mon, 7 Aug 2017 08:57:01 +0000 (10:57 +0200)]
i386/kvm: set tsc_khz before configuring Hyper-V CPUID
Timing-related Hyper-V enlightenments will benefit from knowing the final
tsc_khz value. This commit just moves the code in preparation for further
changes.
Igor Mammedov [Wed, 13 Sep 2017 16:04:57 +0000 (18:04 +0200)]
arm: drop intermediate cpu_model -> cpu type parsing and use cpu type directly
there are 2 use cases to deal with:
1: fixed CPU models per board/soc
2: boards with user configurable cpu_model and fallback to
default cpu_model if user hasn't specified one explicitly
For the 1st
drop intermediate cpu_model parsing and use const cpu type
directly, which replaces:
typename = object_class_get_name(
cpu_class_by_name(TYPE_ARM_CPU, cpu_model))
object_new(typename)
with
object_new(FOO_CPU_TYPE_NAME)
or
cpu_generic_init(BASE_CPU_TYPE, "my cpu model")
with
cpu_create(FOO_CPU_TYPE_NAME)
as result 1st use case doesn't have to invoke not necessary
translation and not needed code is removed.
For the 2nd
1: set default cpu type with MachineClass::default_cpu_type and
2: use generic cpu_model parsing that done before machine_init()
is run and:
2.1: drop custom cpu_model parsing where pattern is:
typename = object_class_get_name(
cpu_class_by_name(TYPE_ARM_CPU, cpu_model))
[parse_features(typename, cpu_model, &err) ]
2.2: or replace cpu_generic_init() which does what
2.1 does + create_cpu(typename) with just
create_cpu(machine->cpu_type)
as result cpu_name -> cpu_type translation is done using
generic machine code one including parsing optional features
if supported/present (removes a bunch of duplicated cpu_model
parsing code) and default cpu type is defined in an uniform way
within machine_class_init callbacks instead of adhoc places
in boadr's machine_init code.
Igor Mammedov [Wed, 13 Sep 2017 16:04:56 +0000 (18:04 +0200)]
pc: use generic cpu_model parsing
define default CPU type in generic way in pc_machine_class_init()
and let common machine code to handle cpu_model parsing
Patch also introduces TARGET_DEFAULT_CPU_TYPE define for 2 purposes:
* make foo_machine_class_init() look uniform on every target
* use define in [bsd|linux]-user targets to pick default
cpu type
Igor Mammedov [Wed, 13 Sep 2017 16:04:55 +0000 (18:04 +0200)]
vl.c: convert cpu_model to cpu type and set of global properties before machine_init()
All machines that support user specified cpu_model either call
cpu_generic_init() or cpu_class_by_name()/CPUClass::parse_features
to parse feature string and to get CPU type to create.
Which leads to code duplication and hard-codding default CPU model
within machine_foo_init() code. Which makes it impossible to
get CPU type before machine_init() is run.
So instead of setting default CPUs models and doing parsing in
target specific machine_foo_init() in various ways, provide
a generic data driven cpu_model parsing before machine_init()
is called.
in follow up per target patches, it will allow to:
* define default CPU type in consistent/generic manner
per machine type and drop custom code that fallbacks
to default if cpu_model is NULL
* drop custom features parsing in targets and do it
in centralized way.
* for cases of
cpu_generic_init(TYPE_BASE/DEFAULT_CPU, "some_cpu")
replace it with
cpu_create(machine->cpu_type) || cpu_create(TYPE_FOO)
depending if CPU type is user settable or not.
not doing useless parsing and clearly documenting where
CPU model is user settable or fixed one.
Patch allows machine subclasses to define default CPU type
per machine class at class_init() time and if that is set
generic code will parse cpu_model into a MachineState::cpu_type
which will be used to create CPUs for that machine instance
and allows gradual per board conversion.
Igor Mammedov [Wed, 13 Sep 2017 16:04:54 +0000 (18:04 +0200)]
cpu: make cpu_generic_init() abort QEMU on error
Almost every user of cpu_generic_init() checks for
returned NULL and then reports failure in a custom way
and aborts process.
Some users assume that call can't fail and don't check
for failure, though they should have checked for it.
In either cases cpu_generic_init() failure is fatal,
so instead of checking for failure and reporting
it various ways, make cpu_generic_init() report
errors in consistent way and terminate QEMU on failure.
Eduardo Habkost [Thu, 24 Aug 2017 19:23:15 +0000 (16:23 -0300)]
hostmem-file: Add "discard-data" option
The new option can be used to indicate that the file contents can
be destroyed and don't need to be flushed to disk when QEMU exits
or when the memory backend object is removed.
Internally, it will trigger a madvise(MADV_REMOVE) call when the
memory backend is removed.
Paolo Bonzini [Tue, 22 Aug 2017 05:08:27 +0000 (07:08 +0200)]
scsi: move non-emulation specific code to scsi/
util/scsi.c includes some SCSI code that is shared by block/iscsi.c and
hw/scsi, but the introduction of the persistent reservation helper
will add many more instances of this. There is also include/block/scsi.h,
which actually is not part of the core block layer.
The persistent reservation manager will also need a home. A scsi/
directory provides one for both the aforementioned shared code and
the PR manager code.
Paolo Bonzini [Tue, 22 Aug 2017 07:31:36 +0000 (09:31 +0200)]
scsi: rename scsi_build_sense to scsi_convert_sense
After introducing the scsi/ subdirectory, there will be a scsi_build_sense
function that is the same as scsi_req_build_sense but without needing
a SCSIRequest. The existing scsi_build_sense function gets in the way,
remove it.
virtio-scsi: Add virtqueue_size parameter allowing virtqueue size to be set.
Since Linux switched to blk-mq as the default in Linux commit 5c279bd9e406 ("scsi: default to scsi-mq"), virtio-scsi LUNs consume
about 10x as much guest kernel memory.
This commit allows you to choose the virtqueue size for each
virtio-scsi-pci controller like this:
-device virtio-scsi-pci,id=scsi,virtqueue_size=16
The default is still 128 as before. Using smaller virtqueue_size
allows many more disks to be added to small memory virtual machines.
For a 1 vCPU, 500 MB, no swap VM I observed:
With scsi-mq enabled (upstream kernel): 175 disks
-"- ditto -"- virtqueue_size=64: 318 disks
-"- ditto -"- virtqueue_size=16: 775 disks
With scsi-mq disabled (kernel before 5c279bd9e406): 1755 disks
Note that to have any effect, this requires a kernel patch:
Joseph Myers [Fri, 11 Aug 2017 14:23:35 +0000 (14:23 +0000)]
target/i386: fix phminposuw in-place operation
The SSE4.1 phminposuw instruction finds the minimum 16-bit element in
the source vector, putting the value of that element in the low 16
bits of the destination vector, the index of that element in the next
three bits and zeroing the rest of the destination. The helper for
this operation fills the destination from high to low, meaning that
when the source and destination are the same register, the minimum
source element can be overwritten before it is copied to the
destination. This patch fixes it to fill the destination from low to
high instead, so the minimum source element is always copied first.
This fixes one gcc test failure in my GCC 6-based testing (and so
concludes the present sequence of patches, as I don't have any further
gcc test failures left in that testing that I attribute to QEMU bugs).
Joseph Myers [Thu, 10 Aug 2017 21:40:41 +0000 (21:40 +0000)]
target/i386: fix pcmpxstrx substring search
One of the cases of the SSE4.2 pcmpestri / pcmpestrm / pcmpistri /
pcmpistrm instructions does a substring search. The implementation of
this case in the pcmpxstrx helper is incorrect. The operation in this
case is a search for a string (argument d to the helper) in another
string (argument s to the helper); if a copy of d at a particular
position would run off the end of s, the resulting output bit should
be 0 whether or not the strings match in the region where they
overlap, but the QEMU implementation was wrongly comparing only up to
the point where s ends and counting it as a match if an initial
segment of d matched a terminal segment of s. Here, "run off the end
of s" means that some byte of d would overlap some byte outside of s;
thus, if d has zero length, it is considered to match everywhere,
including after the end of s. This patch fixes the implementation to
correspond with the proper instruction semantics. This fixes four gcc
test failures in my GCC 6-based testing.
Joseph Myers [Thu, 10 Aug 2017 00:24:23 +0000 (00:24 +0000)]
target/i386: fix packusdw in-place operation
The SSE4.1 packusdw instruction combines source and destination
vectors of signed 32-bit integers into a single vector of unsigned
16-bit integers, with unsigned saturation. When the source and
destination are the same register, this means each 32-bit element of
that register is used twice as an input, to produce two of the 16-bit
output elements, and so if the operation is carried out
element-by-element in-place, no matter what the order in which it is
applied to the elements, the first element's operation will overwrite
some future input. The helper for packssdw avoids this issue by
computing the result in a local temporary and copying it to the
destination at the end; this patch fixes the packusdw helper to do
likewise. This fixes three gcc test failures in my GCC 6-based
testing.
Joseph Myers [Tue, 8 Aug 2017 23:51:29 +0000 (23:51 +0000)]
target/i386: set rip_offset for further SSE instructions
It turns out that my recent fix to set rip_offset when emulating some
SSE4.1 instructions needs generalizing to cover a wider class of
instructions. Specifically, every instruction in the sse_op_table7
table, coming from various instruction set extensions, has an 8-bit
immediate operand that comes after any memory operand, and so needs
rip_offset set for correctness if there is a memory operand that is
rip-relative, and my patch only set it for a subset of those
instructions. This patch moves the rip_offset setting to cover the
wider class of instructions, so fixing 9 further gcc testsuite
failures in my GCC 6-based testing. (I do not know whether there
might be still further classes of instructions missing this setting.)
The SSE4.1 pmovsx* and pmovzx* instructions take packed 1-byte, 2-byte
or 4-byte inputs and sign-extend or zero-extend them to a wider vector
output. The associated helpers for these instructions do the
extension on each element in turn, starting with the lowest. If the
input and output are the same register, this means that all the input
elements after the first have been overwritten before they are read.
This patch makes the helpers extend starting with the highest element,
not the lowest, to avoid such overwriting. This fixes many GCC test
failures (161 in the gcc testsuite in my GCC 6-based testing) when
testing with a default CPU setting enabling those instructions.
* remotes/jnsnow/tags/ide-pull-request:
hw/block/fdc: Convert to realize
hw/ide: Convert DeviceClass init to realize
AHCI: remove DPRINTF macro
AHCI: pretty-print FIS to buffer instead of stderr
AHCI: Rework IRQ constants
AHCI: Replace DPRINTF with trace-events
IDE: replace DEBUG_AIO with trace events
ATAPI: Replace DEBUG_IDE_ATAPI with tracing events
IDE: add tracing for data ports
IDE: Add register hints to tracing
IDE: replace DEBUG_IDE with tracing system
hw/ide/microdrive: Mark the dscm1xxxx device with user_creatable = false
ide: ahci: unparent children buses before freeing their memory
Mao Zhongyi [Mon, 18 Sep 2017 14:05:13 +0000 (22:05 +0800)]
hw/ide: Convert DeviceClass init to realize
Replace init with realize in IDEDeviceClass, which has errp
as a parameter. So all the implementations now use error_setg
instead of error_report for reporting error.
John Snow [Mon, 18 Sep 2017 19:01:26 +0000 (15:01 -0400)]
AHCI: pretty-print FIS to buffer instead of stderr
The current FIS printing routines dump the FIS to screen. adjust this
such that it dumps to buffer instead, then use this ability to have
FIS dump mechanisms via trace-events instead of compiled defines.
John Snow [Mon, 18 Sep 2017 19:01:26 +0000 (15:01 -0400)]
AHCI: Rework IRQ constants
Create a new enum so that we can name the IRQ bits, which will make debugging
them a little nicer if we can print them out. Not handled in this patch, but
this will make it possible to get a nice debug printf detailing exactly which
status bits are set, as it can be multiple at any given time.
As a consequence of this patch, it is no longer possible to set multiple IRQ
codes at once, but nothing was utilizing this ability anyway.
John Snow [Mon, 18 Sep 2017 19:01:26 +0000 (15:01 -0400)]
ATAPI: Replace DEBUG_IDE_ATAPI with tracing events
As part of the ongoing effort to modernize the tracing facilities for
the IDE family of devices, remove PRINTFs in the ATAPI device with
actual tracing events.
Igor Mammedov [Mon, 18 Sep 2017 19:01:25 +0000 (15:01 -0400)]
ide: ahci: unparent children buses before freeing their memory
Fixes read after freeing error reported
https://lists.gnu.org/archive/html/qemu-devel/2017-08/msg04243.html
Message-Id: <59a56959-ca12-ea75-33fa-ff07eba1b090@redhat.com>
ich9-ahci device creates ide buses and attaches them as QOM children
at realize time, however it forgets to properly clean them up
at unrealize time and frees memory containing these children,
with following call-chain:
qdev_device_add()
object_property_set_bool('realized', true)
device_set_realized()
...
pci_qdev_realize() -> pci_ich9_ahci_realize() -> ahci_realize()
...
s->dev = g_new0(AHCIDevice, ports);
...
AHCIDevice *ad = &s->dev[i];
ide_bus_new(&ad->port, sizeof(ad->port), qdev, i, 1);
^^^ creates bus in memory allocated by above gnew()
and adds it as child propety to ahci device
...
hotplug_handler_plug(); -> goto post_realize_fail;
pci_qdev_unrealize() -> pci_ich9_uninit() -> ahci_uninit()
...
g_free(s->dev);
^^^ free memory that holds children busses
return with error from device_set_realized()
As result later when qdev_device_add() tries to unparent ich9-ahci
after failed device_set_realized(),
object_unparent() -> object_property_del_child()
iterates over existing QOM children including buses added by
ide_bus_new() and tries to unparent them, which causes access to
freed memory where they where located.
Kevin Wolf [Mon, 18 Sep 2017 05:25:24 +0000 (07:25 +0200)]
qemu.py: Fix syntax error
Python requires parentheses around multiline expression. This fixes the
breakage of all Python-based qemu-iotests cases that was introduced in
commit dab91d9aa0.
Peter Maydell [Sun, 17 Sep 2017 15:24:48 +0000 (16:24 +0100)]
Merge remote-tracking branch 'remotes/rth/tags/pull-tcg-20170917' into staging
tcg queued patches
# gpg: Signature made Sun 17 Sep 2017 16:03:28 BST
# gpg: using RSA key 0x64DF38E8AF7E215F
# gpg: Good signature from "Richard Henderson <[email protected]>"
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg: It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A 05C0 64DF 38E8 AF7E 215F
* remotes/rth/tags/pull-tcg-20170917:
tcg/mips: Fully convert tcg_target_op_def
tcg/sparc: Fully convert tcg_target_op_def
tcg/ppc: Fully convert tcg_target_op_def
tcg/arm: Fully convert tcg_target_op_def
tcg/aarch64: Fully convert tcg_target_op_def
tcg: Fix types in tcg_regset_{set,reset}_reg
tcg: Remove tcg_regset_set32
tcg: Remove tcg_regset_{or,and,andnot,not}
tcg: Remove tcg_regset_set
tcg: Remove tcg_regset_clear
tcg: Add tcg_op_supported
accel/tcg: move USER code to user-exec.c
accel/tcg: move atomic_template.h to accel/tcg/
accel/tcg: move tcg-runtime to accel/tcg/
accel/tcg: move user-exec to accel/tcg/
accel/tcg: move softmmu_template.h to accel/tcg/
tcg/ppc: disable atomic write check on ppc32
projects / qemu.git / log