Paul Lai [Wed, 26 Jun 2019 16:21:29 +0000 (09:21 -0700)]
i386: Introduce SnowRidge CPU model
SnowRidge CPU supports Accelerator Infrastrcture Architecture (MOVDIRI,
MOVDIR64B), CLDEMOTE and SPLIT_LOCK_DISABLE.
MOVDIRI, MOVDIR64B, and CLDEMOTE are found via CPUID.
The availability of SPLIT_LOCK_DISABLE is check via msr access
References can be found in either:
https://software.intel.com/en-us/articles/intel-sdm
https://software.intel.com/en-us/download/intel-architecture-instruction-set-extensions-and-future-features-programming-reference
Eduardo Habkost [Sat, 8 Jun 2019 23:34:47 +0000 (20:34 -0300)]
qmp: Add deprecation information to query-machines
Export machine type deprecation status through the query-machines
QMP command. With this, libvirt and management software will be
able to show this information to users and/or suggest changes to
VM configuration to avoid deprecated machines.
Like Xu [Thu, 20 Jun 2019 05:45:25 +0000 (13:45 +0800)]
vl.c: Add -smp, dies=* command line support and update doc
For PC target, users could configure the number of dies per one package
via command line with this patch, such as "-smp dies=2,cores=4".
The parsing rules of new cpu-topology model obey the same restrictions/logic
as the legacy socket/core/thread model especially on missing values computing.
Like Xu [Thu, 20 Jun 2019 05:45:24 +0000 (13:45 +0800)]
machine: Refactor smp_parse() in vl.c as MachineClass::smp_parse()
To make smp_parse() more flexible and expansive, a smp_parse function
pointer is added to MachineClass that machine types could override.
The generic smp_parse() code in vl.c is moved to hw/core/machine.c, and
become the default implementation of MachineClass::smp_parse. A PC-specific
function called pc_smp_parse() has been added to hw/i386/pc.c, which in
this patch changes nothing against the default one .
Like Xu [Thu, 20 Jun 2019 05:45:23 +0000 (13:45 +0800)]
target/i386: Add CPUID.1F generation support for multi-dies PCMachine
The CPUID.1F as Intel V2 Extended Topology Enumeration Leaf would be
exposed if guests want to emulate multiple software-visible die within
each package. Per Intel's SDM, the 0x1f is a superset of 0xb, thus they
can be generated by almost same code as 0xb except die_offset setting.
If the number of dies per package is greater than 1, the cpuid_min_level
would be adjusted to 0x1f regardless of whether the host supports CPUID.1F.
Likewise, the CPUID.1F wouldn't be exposed if env->nr_dies < 2.
Roman Kagan [Tue, 18 Jun 2019 11:07:06 +0000 (11:07 +0000)]
i386: make 'hv-spinlocks' a regular uint32 property
X86CPU.hv-spinlocks is a uint32 property that has a special setter
validating the value to be no less than 0xFFF and no bigger than
UINT_MAX. The latter check is redundant; as for the former, there
appears to be no reason to prohibit the user from setting it to a lower
value.
So nuke the dedicated getter/setter pair and convert 'hv-spinlocks' to a
regular uint32 property.
Eduardo Habkost [Sat, 15 Jun 2019 20:05:05 +0000 (17:05 -0300)]
i386: Fix signedness of hyperv_spinlock_attempts
The current default value for hv-spinlocks is 0xFFFFFFFF (meaning
"never retry"). However, the value is stored as a signed
integer, making the getter of the hv-spinlocks QOM property
return -1 instead of 0xFFFFFFFF.
Fix this by changing the type of X86CPU::hyperv_spinlock_attempts
to uint32_t. This has no visible effect to guest operating
systems, affecting just the behavior of the QOM getter.
Eduardo Habkost [Tue, 11 Jun 2019 20:54:20 +0000 (17:54 -0300)]
i386: Don't print warning if phys-bits was set automatically
If cpu->host_phys_bits_limit is set, QEMU will make
cpu->phys_bits be lower than host_phys_bits on some cases. This
triggers a warning that was supposed to be printed only if
phys-bits was explicitly set in the command-line.
Reorder the code so the value of cpu->phys_bits is validated
before the cpu->host_phys_bits handling. This will avoid
unexpected warnings when cpu->host_phys_bits_limit is set.
Igor Mammedov [Wed, 26 Jun 2019 07:42:28 +0000 (03:42 -0400)]
deprecate -mem-path fallback to anonymous RAM
Fallback might affect guest or worse whole host performance
or functionality if backing file were used to share guest RAM
with another process.
Patch deprecates fallback so that we could remove it in future
and ensure that QEMU will provide expected behavior and fail if
it can't use user provided backing file.
Igor Mammedov [Thu, 30 May 2019 08:33:19 +0000 (10:33 +0200)]
numa: deprecate implict memory distribution between nodes
Implicit RAM distribution between nodes has exactly the same issues as:
"numa: deprecate 'mem' parameter of '-numa node' option"
only with QEMU being the user that's 'adding' 'mem' parameter.
Deprecate it, to get it out of the way so that we could consolidate
guest RAM allocation using memory backends making it consistent and
possibly later on transition to using memory devices instead of
adhoc memory mapping for the initial RAM.
Igor Mammedov [Thu, 30 May 2019 08:33:18 +0000 (10:33 +0200)]
numa: deprecate 'mem' parameter of '-numa node' option
The parameter allows to configure fake NUMA topology where guest
VM simulates NUMA topology but not actually getting performance
benefits from it. The same or better results could be achieved
using 'memdev' parameter.
Beside of unpredictable performance, '-numa node.mem' option has
other issues when it's used with combination of -mem-path +
+ -mem-prealloc + memdev backends (pc-dimm), breaking binding of
memdev backends since mem-path/mem-prealloc are global and affect
the most of RAM allocations.
It's possible to make memdevs and global -mem-path/mem-prealloc
to play nicely together but that will just complicate already
complicated code and add unobious ways it could break on 2
different memmory allocation pathes and their combinations.
Instead of it, consolidate all guest RAM allocation over memdev
which still allows to create fake NUMA configurations if desired
and leaves one simplifyed code path to consider when it comes
to guest RAM allocation.
To achieve desired simplification deprecate 'mem' parameter as its
ad-hoc partitioning of initial RAM MemoryRegion can't be translated
to memdev based backend transparently to users and in compatible
manner (migration wise).
Later down the road that will allow to consolidate means of how
guest RAM is allocated and would permit us to clean up quite
a bit memory allocations and numa code, leaving only 'memdev'
implementation in place.
Igor Mammedov [Mon, 10 Jun 2019 13:10:07 +0000 (15:10 +0200)]
machine: show if CLI option '-numa node,mem' is supported in QAPI schema
Legacy '-numa node,mem' option has a number of issues and mgmt often
defaults to it. Unfortunately it's no possible to replace it with
an alternative '-numa memdev' without breaking migration compatibility.
What's possible though is to deprecate it, keeping option working with
old machine types only.
In order to help users to find out if being deprecated CLI option
'-numa node,mem' is still supported by particular machine type, add new
"numa-mem-supported" property to output of query-machines.
"numa-mem-supported" is set to 'true' for machines that currently support
NUMA, but it will be flipped to 'false' later on, once deprecation period
expires and kept 'true' only for old machine types that used to support
the legacy option so it won't break existing configuration that are using
it.
Like Xu [Wed, 12 Jun 2019 08:40:59 +0000 (16:40 +0800)]
i386: Update new x86_apicid parsing rules with die_offset support
In new sockets/dies/cores/threads model, the apicid of logical cpu could
imply die level info of guest cpu topology thus x86_apicid_from_cpu_idx()
need to be refactored with #dies value, so does apicid_*_offset().
To keep semantic compatibility, the legacy pkg_offset which helps to
generate CPUIDs such as 0x3 for L3 cache should be mapping to die_offset.
Like Xu [Wed, 12 Jun 2019 08:40:58 +0000 (16:40 +0800)]
i386/cpu: Consolidate die-id validity in smp context
The field die_id (default as 0) and has_die_id are introduced to X86CPU.
Following the legacy smp check rules, the die_id validity is added to
the same contexts as leagcy smp variables such as hmp_hotpluggable_cpus(),
machine_set_cpu_numa_node(), cpu_slot_to_string() and pc_cpu_pre_plug().
Like Xu [Wed, 12 Jun 2019 08:40:57 +0000 (16:40 +0800)]
hw/i386: Adjust nr_dies with configured smp_dies for PCMachine
To support multiple dies configuration on PCMachine, the best place to
set CPUX86State->nr_dies with requested PCMachineState->smp_dies is in
pc_new_cpu() and pc_cpu_pre_plug(). Refactoring pc_new_cpu() is applied
and redundant parameter "const char *typename" would be removed.
Like Xu [Wed, 12 Jun 2019 08:40:56 +0000 (16:40 +0800)]
i386: Add die-level cpu topology to x86CPU on PCMachine
The die-level as the first PC-specific cpu topology is added to the leagcy
cpu topology model, which has one die per package implicitly and only the
numbers of sockets/cores/threads are configurable.
In the new model with die-level support, the total number of logical
processors (including offline) on board will be calculated as:
#cpus = #sockets * #dies * #cores * #threads
and considering compatibility, the default value for #dies would be
initialized to one in x86_cpu_initfn() and pc_machine_initfn().
Like Xu [Sat, 18 May 2019 20:54:28 +0000 (04:54 +0800)]
vl.c: Replace smp global variables with smp machine properties
The global smp variables in vl.c are completely replaced with machine properties.
Form this commit, the smp_cpus/smp_cores/smp_threads/max_cpus are deprecated
and only machine properties within MachineState are fully applied and enabled.
Like Xu [Sat, 18 May 2019 20:54:27 +0000 (04:54 +0800)]
hw: Replace global smp variables with MachineState for all remaining archs
The global smp variables in alpha/hppa/mips/openrisc/sparc*/xtensa codes
are replaced with smp properties from MachineState.
A local variable of the same name would be introduced in the declaration
phase if it's used widely in the context OR replace it on the spot if it's
only used once. No semantic changes.
Like Xu [Sat, 18 May 2019 20:54:26 +0000 (04:54 +0800)]
hw/arm: Replace global smp variables with machine smp properties
The global smp variables in arm are replaced with smp machine properties.
The init_cpus() and *_create_rpu() are refactored to pass MachineState.
A local variable of the same name would be introduced in the declaration
phase if it's used widely in the context OR replace it on the spot if it's
only used once. No semantic changes.
Like Xu [Sat, 18 May 2019 20:54:25 +0000 (04:54 +0800)]
hw/i386: Replace global smp variables with machine smp properties
The global smp variables in i386 are replaced with smp machine properties.
To avoid calling qdev_get_machine() as much as possible, some related funtions
for acpi data generations are refactored. No semantic changes.
A local variable of the same name would be introduced in the declaration
phase if it's used widely in the context OR replace it on the spot if it's
only used once. No semantic changes.
Like Xu [Sat, 18 May 2019 20:54:24 +0000 (04:54 +0800)]
hw/s390x: Replace global smp variables with machine smp properties
The global smp variables in s390x are replaced with smp machine properties.
A local variable of the same name would be introduced in the declaration
phase if it's used widely in the context OR replace it on the spot if it's
only used once. No semantic changes.
Like Xu [Sat, 18 May 2019 20:54:23 +0000 (04:54 +0800)]
hw/riscv: Replace global smp variables with machine smp properties
The global smp variables in riscv are replaced with smp machine properties.
A local variable of the same name would be introduced in the declaration
phase if it's used widely in the context OR replace it on the spot if it's
only used once. No semantic changes.
Like Xu [Sat, 18 May 2019 20:54:22 +0000 (04:54 +0800)]
hw/ppc: Replace global smp variables with machine smp properties
The global smp variables in ppc are replaced with smp machine properties.
A local variable of the same name would be introduced in the declaration
phase if it's used widely in the context OR replace it on the spot if it's
only used once. No semantic changes.
Like Xu [Sat, 18 May 2019 20:54:21 +0000 (04:54 +0800)]
general: Replace global smp variables with smp machine properties
Basically, the context could get the MachineState reference via call
chains or unrecommended qdev_get_machine() in !CONFIG_USER_ONLY mode.
A local variable of the same name would be introduced in the declaration
phase out of less effort OR replace it on the spot if it's only used
once in the context. No semantic changes.
Like Xu [Sat, 18 May 2019 20:54:20 +0000 (04:54 +0800)]
machine: Refactor smp-related call chains to pass MachineState
To get rid of the global smp_* variables we're currently using, it's recommended
to pass MachineState in the list of incoming parameters for functions that use
global smp variables, thus some redundant parameters are dropped. It's applied
for legacy smbios_*(), *_machine_reset(), hot_add_cpu() and mips *_create_cpu().
Like Xu [Sat, 18 May 2019 20:54:19 +0000 (04:54 +0800)]
hw/boards: Add struct CpuTopology to MachineState
The cpu topology property CpuTopology is added to the MachineState
and its members are initialized with the leagcy global smp variables.
From this commit, the code in the system emulation mode is supposed to
use cpu topology variables from MachineState instead of the global ones
defined in vl.c and there is no semantic change.
* remotes/kraxel/tags/vga-20190705-pull-request:
ati-vga: Fix setting offset together with pitch for r128pro
ati-vga: Fix reverse bit blts
ati-vga: Fix frame buffer endianness for big endian target
ati-vga: Improve readability of ati_2d_blt function
Peter Maydell [Fri, 5 Jul 2019 10:21:29 +0000 (11:21 +0100)]
Merge remote-tracking branch 'remotes/stsquad/tags/pull-testing-next-050719-3' into staging
Various testing fixes:
- tests/vm updates and clean-ups
- tests/vm serial autobuild on host (-netbsd v3)
- ensure MacOS builds do "brew update"
- ensure we test --static user builds
- fix hyperv compile failure
- fix missing var warning for OpenBSD (v2)
This brings my testing back to green on all CI services. Please note
the BSD installs will throw out some warnings during the setup phase.
They shouldn't re-occur once the images are built. NetBSD has been
dropped for now given slow install issues.
# gpg: Signature made Fri 05 Jul 2019 11:15:21 BST
# gpg: using RSA key 6685AE99E75167BCAFC8DF35FBD0DB095A9E2A44
# gpg: Good signature from "Alex Bennée (Master Work Key) <[email protected]>" [full]
# Primary key fingerprint: 6685 AE99 E751 67BC AFC8 DF35 FBD0 DB09 5A9E 2A44
* remotes/stsquad/tags/pull-testing-next-050719-3:
migration: move port_attr inside CONFIG_LINUX
target/i386: fix feature check in hyperv-stub.c
Makefile: Rename the 'vm-test' target as 'vm-help'
.travis.yml: force a brew update for MacOS builds
.travis.yml: default the --disable-system build to --static
tests/vm: ubuntu.i386: apt proxy setup
tests/vm: fedora autoinstall, using serial console
tests/vm: freebsd autoinstall, using serial console
tests/vm: openbsd autoinstall, using serial console
tests/vm: serial console support helpers
tests/vm: add vm-boot-{ssh,serial}-<guest> targets
tests/vm: proper guest shutdown
tests/vm: run test builds on snapshot
tests/vm: use ssh with pty unconditionally
tests/vm: send proxy environment variables over ssh
tests/vm: add source repos on ubuntu.i386
tests/vm: pin ubuntu.i386 image
tests/vm: avoid image presence check and removal
tests/vm: avoid extra compressed image copy
Peter Maydell [Fri, 5 Jul 2019 08:51:50 +0000 (09:51 +0100)]
Merge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging
virtio, pc, pci: features, fixes, cleanups
virtio-pmem support.
libvhost user mq support.
A bunch of fixes all over the place.
Signed-off-by: Michael S. Tsirkin <[email protected]>
# gpg: Signature made Thu 04 Jul 2019 22:00:49 BST
# gpg: using RSA key 281F0DB8D28D5469
# gpg: Good signature from "Michael S. Tsirkin <[email protected]>" [full]
# gpg: aka "Michael S. Tsirkin <[email protected]>" [full]
# Primary key fingerprint: 0270 606B 6F3C DF3D 0B17 0970 C350 3912 AFBE 8E67
# Subkey fingerprint: 5D09 FD08 71C8 F85B 94CA 8A0D 281F 0DB8 D28D 5469
* remotes/mst/tags/for_upstream: (22 commits)
docs: avoid vhost-user-net specifics in multiqueue section
libvhost-user: implement VHOST_USER_PROTOCOL_F_MQ
libvhost-user: support many virtqueues
libvhost-user: add vmsg_set_reply_u64() helper
pc: Move compat_apic_id_mode variable to PCMachineClass
virtio: Don't change "started" flag on virtio_vmstate_change()
virtio: Make sure we get correct state of device on handle_aio_output()
virtio: Set "start_on_kick" on virtio_set_features()
virtio: Set "start_on_kick" for legacy devices
virtio: add "use-started" property
virtio-pci: fix missing device properties
pc: Support for virtio-pmem-pci
numa: Handle virtio-pmem in NUMA stats
hmp: Handle virtio-pmem when printing memory device infos
virtio-pci: Proxy for virtio-pmem
virtio-pmem: sync linux headers
virtio-pci: Allow to specify additional interfaces for the base type
virtio-pmem: add virtio device
pcie: minor cleanups for slot control/status
pcie: work around for racy guest init
...
ati-vga: Fix setting offset together with pitch for r128pro
Turns out my last fix to this broke one case for Rage 128 Pro so
revert that part of previous patch. This now fixes the remaining
rendering problems for MorphOS which now can produce picture with
-device ati-vga (although it may not be optimised yet and video
overlay emulation is still known to be missing).
The pixman library only supports blts with left to right, top to
bottom order but the ATI VGA engine can also do different directions.
Fix support for these via a temporary buffer for now. This fixes
rendering issues related to such blts (such as moving windows) but
some other glitches still remain.
Stefan Hajnoczi [Wed, 26 Jun 2019 07:48:15 +0000 (08:48 +0100)]
docs: avoid vhost-user-net specifics in multiqueue section
The "Multiple queue support" section makes references to vhost-user-net
"queue pairs". This is confusing for two reasons:
1. This actually applies to all device types, not just vhost-user-net.
2. VHOST_USER_GET_QUEUE_NUM returns the number of virtqueues, not the
number of queue pairs.
Reword the section so that the vhost-user-net specific part is relegated
to the very end: we acknowledge that vhost-user-net historically
automatically enabled the first queue pair.
Stefan Hajnoczi [Wed, 26 Jun 2019 07:48:14 +0000 (08:48 +0100)]
libvhost-user: implement VHOST_USER_PROTOCOL_F_MQ
Existing vhost-user device backends, including vhost-user-scsi and
vhost-user-blk, support multiqueue but libvhost-user currently does not
advertise this.
VHOST_USER_PROTOCOL_F_MQ enables the VHOST_USER_GET_QUEUE_NUM request
needed for a vhost-user master to query the number of queues. For
example, QEMU's vhost-user-net master depends on
VHOST_USER_PROTOCOL_F_MQ for multiqueue.
If you're wondering how any device backend with more than one virtqueue
functions today, it's because device types with a fixed number of
virtqueues do not require querying the number of queues. Therefore the
vhost-user master for vhost-user-input with 2 virtqueues, for example,
doesn't actually depend on VHOST_USER_PROTOCOL_F_MQ. It just enables
virtqueues 0 and 1 without asking.
Stefan Hajnoczi [Wed, 26 Jun 2019 07:48:13 +0000 (08:48 +0100)]
libvhost-user: support many virtqueues
Currently libvhost-user is hardcoded to at most 8 virtqueues. The
device backend should decide the number of virtqueues, not
libvhost-user. This is important for multiqueue device backends where
the guest driver needs an accurate number of virtqueues.
This change breaks libvhost-user and libvhost-user-glib API stability.
There is no stability guarantee yet, so make this change now and update
all in-tree library users.
This patch touches up vhost-user-blk, vhost-user-gpu, vhost-user-input,
vhost-user-scsi, and vhost-user-bridge. If the device has a fixed
number of queues that exact number is used. Otherwise the previous
default of 8 virtqueues is used.
vu_init() and vug_init() can now fail if malloc() returns NULL. I
considered aborting with an error in libvhost-user but it should be safe
to instantiate new vhost-user instances at runtime without risk of
terminating the process. Therefore callers need to handle the vu_init()
failure now.
vhost-user-blk and vhost-user-scsi duplicate virtqueue index checks that
are already performed by libvhost-user. This code would need to be
modified to use max_queues but remove it completely instead since it's
redundant.
Stefan Hajnoczi [Wed, 26 Jun 2019 07:48:12 +0000 (08:48 +0100)]
libvhost-user: add vmsg_set_reply_u64() helper
The VhostUserMsg request is reused as the reply by message processing
functions. This is risky since request fields may corrupt the reply if
the vhost-user message handler function forgets to re-initialize them.
Changing this practice would be very invasive but we can introduce a
helper function to make u64 payload replies safe. This also eliminates
code duplication in message processing functions.
Xie Yongji [Wed, 26 Jun 2019 02:31:30 +0000 (10:31 +0800)]
virtio: Don't change "started" flag on virtio_vmstate_change()
We will call virtio_set_status() on virtio_vmstate_change().
The "started" flag should not be changed in this case. Otherwise,
we may get an incorrect value when we set "started" flag but
not set DRIVER_OK in source VM.
Xie Yongji [Wed, 26 Jun 2019 02:31:28 +0000 (10:31 +0800)]
virtio: Set "start_on_kick" on virtio_set_features()
The guest feature is not set correctly on virtio_reset() and
virtio_init(). So we should not use it to set "start_on_kick" at that
point. This patch set "start_on_kick" on virtio_set_features() instead.
Xie Yongji [Wed, 26 Jun 2019 02:31:26 +0000 (10:31 +0800)]
virtio: add "use-started" property
In order to avoid migration issues, we introduce a "use-started"
property to the base virtio device to indicate whether use
"started" flag or not. This property will be true by default and
set to false when machine type <= 4.0.
Since commit a4ee4c8baa37154 ("virtio: Helper for registering virtio
device types"), virtio-gpu-pci, virtio-vga, and virtio-crypto-pci lost
some properties: "ioeventfd" and "vectors". This may cause various
issues, such as failing migration or invalid properties.
Since those VirtioPCI devices do not have a base name, their class are
initialized with virtio_pci_generic_base_class_init(). However, if the
VirtioPCIDeviceTypeInfo provided a class_init which sets dc->props,
the properties were overwritten by virtio_pci_generic_class_init().
Instead, introduce an intermediary base-type to register the generic
properties.
Override the device hotplug handler to properly handle the memory device
part via virtio-pmem-pci callbacks from the machine hotplug handler and
forward to the actual PCI bus hotplug handler.
As PCI hotplug has not been properly factored out into hotplug handlers,
most magic is performed in the (un)realize functions. Also some PCI host
buses don't have a PCI hotplug handler at all yet, just to be sure that
we alway have a hotplug handler on x86, add a simple error check.
Unlocking virtio-pmem will unlock virtio-pmem-pci.
Pankaj Gupta [Wed, 19 Jun 2019 09:49:04 +0000 (15:19 +0530)]
virtio-pci: Proxy for virtio-pmem
We need a proxy device for virtio-pmem, and this device has to be the
actual memory device so we can cleanly hotplug it.
Forward memory device class functions either to the actual device or use
properties of the virtio-pmem device to implement these in the proxy.
virtio-pmem will only be compiled for selected, supported architectures
(that can deal with virtio/pci devices being memory devices). An
architecture that is prepared for that can simply enable
CONFIG_VIRTIO_PMEM to make it work.
As not all architectures support memory devices (and CONFIG_VIRTIO_PMEM
will be enabled per supported architecture), we have to move the PCI proxy
to a separate file.
Pankaj Gupta [Wed, 19 Jun 2019 09:49:03 +0000 (15:19 +0530)]
virtio-pmem: sync linux headers
Add linux headers for virtio pmem. These are not yet upstream - include
them temporarily as merge window in which this is supposed to be is
coming up shortly. If virtio-pmem ends up not being merged
then this will be reverted and accordingly virtio-pmem dropped.
The reason was the conversion of cpu->hyperv_synic to
cpu->hyperv_synic_kvm_only although the rest of the patch introduces a
feature checking mechanism. So I've fixed the KVM_EXIT_HYPERV_SYNIC in
hyperv-stub to do the same feature check as in the real hyperv.c
Alex Bennée [Mon, 17 Jun 2019 15:35:22 +0000 (16:35 +0100)]
.travis.yml: default the --disable-system build to --static
It's fairly common to build qemu-user binaries with --static linking
so the binary can be copied around without libraries. Enable --static
in the default qemu-user build to cover this.
There are other qemu-user builds that use dynamic linking so they
should catch any problems there.
Gerd Hoffmann [Mon, 17 Jun 2019 04:38:57 +0000 (06:38 +0200)]
tests/vm: fedora autoinstall, using serial console
Download the install iso and prepare the image locally. Install to
disk, using the serial console. Create qemu user, configure ssh login.
Install packages needed for qemu builds.
Yes, we have docker images for fedora. But for trouble-shooting it
might be helpful to have a vm too. When vm builds fail you can use
it to figure whenever the vm setup or the guest os is the problem.
Gerd Hoffmann [Mon, 17 Jun 2019 04:38:55 +0000 (06:38 +0200)]
tests/vm: freebsd autoinstall, using serial console
Instead of fetching the prebuilt image from patchew download the install
iso and prepare the image locally. Install to disk, using the serial
console. Create qemu user, configure ssh login. Install packages
needed for qemu builds.
Note that freebsd package downloads are delivered as non-cachable
content, so I had to configure squid with "ignore-no-store
ignore-private ignore-reload" for pkgmir.geo.freebsd.org to make the
caching actually work.
Gerd Hoffmann [Mon, 17 Jun 2019 04:38:54 +0000 (06:38 +0200)]
tests/vm: openbsd autoinstall, using serial console
Instead of fetching the prebuilt image from patchew download the install
iso and prepare the image locally. Install to disk, using the serial
console. Create qemu user, configure ssh login. Install packages
needed for qemu builds.
Gerd Hoffmann [Mon, 17 Jun 2019 04:38:51 +0000 (06:38 +0200)]
tests/vm: proper guest shutdown
When not running in snapshot mode ask the guest to poweroff and wait for
this to finish instead of simply quitting qemu, so the guest can flush
pending updates to disk.
Gerd Hoffmann [Mon, 17 Jun 2019 04:38:48 +0000 (06:38 +0200)]
tests/vm: send proxy environment variables over ssh
Packages are fetched via proxy that way, if configured on the host.
That might be required to pass firewalls, and it allows to route
package downloads through a caching proxy server.
Needs AcceptEnv setup in sshd_config on the guest side to work.
Cleber Rosa [Thu, 13 Jun 2019 13:07:17 +0000 (09:07 -0400)]
tests/vm: pin ubuntu.i386 image
It's a good practice to always have the same components used in tests.
According to:
https://cloud-images.ubuntu.com/releases/16.04/
New images are released from time to time, and the "release/"
directory points to the latest release. Let's pin to the latest
available version, and while at it, set a hash for verification.
Cleber Rosa [Thu, 13 Jun 2019 13:07:15 +0000 (09:07 -0400)]
tests/vm: avoid extra compressed image copy
The image copy is only really needed because xz doesn't know to
properly decompress a file not named properly. Instead of
decompressing to stdout, and having to rely on a shell, let's just
create a link instead of copying the file.
Peter Maydell [Thu, 4 Jul 2019 16:32:24 +0000 (17:32 +0100)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20190704-1' into staging
target-arm queue:
* more code-movement to separate TCG-only functions into their own files
* Correct VMOV_imm_dp handling of short vectors
* Execute Thumb instructions when their condbits are 0xf
* armv7m_systick: Forbid non-privileged accesses
* Use _ra versions of cpu_stl_data() in v7M helpers
* v8M: Check state of exception being returned from
* v8M: Forcibly clear negative-priority exceptions on deactivate
* remotes/pmaydell/tags/pull-target-arm-20190704-1:
target/arm: Correct VMOV_imm_dp handling of short vectors
target/arm: Execute Thumb instructions when their condbits are 0xf
hw/timer/armv7m_systick: Forbid non-privileged accesses
target/arm: Use _ra versions of cpu_stl_data() in v7M helpers
target/arm: v8M: Check state of exception being returned from
arm v8M: Forcibly clear negative-priority exceptions on deactivate
target/arm/helper: Move M profile routines to m_helper.c
target/arm: Restrict semi-hosting to TCG
target/arm: Move debug routines to debug_helper.c
Peter Maydell [Thu, 4 Jul 2019 16:14:44 +0000 (17:14 +0100)]
target/arm: Correct VMOV_imm_dp handling of short vectors
Coverity points out (CID 1402195) that the loop in trans_VMOV_imm_dp()
that iterates over the destination registers in a short-vector VMOV
accidentally throws away the returned updated register number
from vfp_advance_dreg(). Add the missing assignment. (We got this
correct in trans_VMOV_imm_sp().)
Peter Maydell [Thu, 4 Jul 2019 16:14:44 +0000 (17:14 +0100)]
target/arm: Execute Thumb instructions when their condbits are 0xf
Thumb instructions in an IT block are set up to be conditionally
executed depending on a set of condition bits encoded into the IT
bits of the CPSR/XPSR. The architecture specifies that if the
condition bits are 0b1111 this means "always execute" (like 0b1110),
not "never execute"; we were treating it as "never execute". (See
the ConditionHolds() pseudocode in both the A-profile and M-profile
Arm ARM.)
This is a bit of an obscure corner case, because the only legal
way to get to an 0b1111 set of condbits is to do an exception
return which sets the XPSR/CPSR up that way. An IT instruction
which encodes a condition sequence that would include an 0b1111 is
UNPREDICTABLE, and for v8A the CONSTRAINED UNPREDICTABLE choices
for such an IT insn are to NOP, UNDEF, or treat 0b1111 like 0b1110.
Add a comment noting that we take the latter option.
Like most of the v7M memory mapped system registers, the systick
registers are accessible to privileged code only and user accesses
must generate a BusFault. We implement that for registers in
the NVIC proper already, but missed it for systick since we
implement it as a separate device. Correct the omission.
Peter Maydell [Thu, 4 Jul 2019 16:14:44 +0000 (17:14 +0100)]
target/arm: Use _ra versions of cpu_stl_data() in v7M helpers
In the various helper functions for v7M/v8M instructions, use
the _ra versions of cpu_stl_data() and friends. Otherwise we
may get wrong behaviour or an assert() due to not being able
to locate the TB if there is an exception on the memory access
or if it performs an IO operation when in icount mode.
Peter Maydell [Thu, 4 Jul 2019 16:14:44 +0000 (17:14 +0100)]
target/arm: v8M: Check state of exception being returned from
In v8M, an attempt to return from an exception which is not
active is an illegal exception return. For this purpose,
exceptions which can configurably target either Secure or
NonSecure are not considered to be active if they are
configured for the opposite security state for the one
we're trying to return from (eg attempt to return from
an NS NMI but NMI targets Secure). In the pseudocode this
is handled by IsActiveForState().
Detect this case rather than counting an active exception
possibly of the wrong security state as being sufficient.
Peter Maydell [Thu, 4 Jul 2019 16:14:43 +0000 (17:14 +0100)]
arm v8M: Forcibly clear negative-priority exceptions on deactivate
To prevent execution priority remaining negative if the guest
returns from an NMI or HardFault with a corrupted IPSR, the
v8M interrupt deactivation process forces the HardFault and NMI
to inactive based on the current raw execution priority,
even if the interrupt the guest is trying to deactivate
is something else. In the pseudocode this is done in the
Deactivate() function.
* remotes/armbru/tags/pull-build-2019-07-02-v2:
Makefile: Reuse all's recursion machinery for clean and install
Makefile: Rename targets for make recursion
Makefile: Drop bogus cleaning of $(ALL_SUBDIRS)/qemu-options.def
Makefile: Remove code to smooth transition to config.status
* remotes/ehabkost/tags/python-next-pull-request:
Deprecate Python 2 support
machine.py: minor delinting
python/qemu: split QEMUMachine out from underneath __init__.py
qmp: make qmp-shell work with python3
Peter Maydell [Thu, 4 Jul 2019 10:09:19 +0000 (11:09 +0100)]
Merge remote-tracking branch 'remotes/palmer/tags/riscv-for-master-4.1-sf1-v3' into staging
RISC-V Patches for the 4.1 Soft Freeze, Part 2 v3
This pull request contains a handful of patches that I'd like to target
for the 4.1 soft freeze. There are a handful of new features:
* Support for the 1.11.0, the latest privileged specification.
* Support for reading and writing the PRCI registers.
* Better control over the ISA of the target machine.
* Support for the cpu-topology device tree node.
Additionally, there are a handful of bug fixes including:
* Load reservations are now broken by both store conditional and by
scheduling, which fixes issues with parallel applications.
* Various fixes to the PMP implementation.
* Fixes to the 32-bit linux-user syscall ABI.
* Various fixes for instruction decodeing.
* A fix to the PCI device tree "bus-range" property.
This boots 32-bit and 64-bit OpenEmbedded.
Changes since v2 [riscv-for-master-4.1-sf1-v2]:
* Dropped OpenSBI.
Changes since v1 [riscv-for-master-4.1-sf1]:
* Contains a fix to the sifive_u OpenSBI integration.
# gpg: Signature made Wed 03 Jul 2019 09:39:09 BST
# gpg: using RSA key 00CE76D1834960DFCE886DF8EF4CA1502CCBAB41
# gpg: issuer "[email protected]"
# gpg: Good signature from "Palmer Dabbelt <[email protected]>" [unknown]
# gpg: aka "Palmer Dabbelt <[email protected]>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 00CE 76D1 8349 60DF CE88 6DF8 EF4C A150 2CCB AB41
* remotes/palmer/tags/riscv-for-master-4.1-sf1-v3: (32 commits)
hw/riscv: Extend the kernel loading support
hw/riscv: Add support for loading a firmware
hw/riscv: Split out the boot functions
riscv: sifive_u: Update the plic hart config to support multicore
riscv: sifive_u: Do not create hard-coded phandles in DT
disas/riscv: Fix `rdinstreth` constraint
disas/riscv: Disassemble reserved compressed encodings as illegal
riscv: virt: Add cpu-topology DT node.
RISC-V: Update syscall list for 32-bit support.
RISC-V: Clear load reservations on context switch and SC
RISC-V: Add support for the Zicsr extension
RISC-V: Add support for the Zifencei extension
target/riscv: Add support for disabling/enabling Counters
target/riscv: Remove user version information
target/riscv: Require either I or E base extension
qemu-deprecated.texi: Deprecate the RISC-V privledge spec 1.09.1
target/riscv: Set privledge spec 1.11.0 as default
target/riscv: Add the mcountinhibit CSR
target/riscv: Add the privledge spec version 1.11.0
target/riscv: Restructure deprecatd CPUs
...
* remotes/mcayland/tags/qemu-sparc-20190702:
sunhme: ensure that RX descriptor ring overflow is indicated to client driver
sunhme: fix return values from sunhme_receive() during receive packet processing
sunhme: flush any queued packets when HME_MAC_RXCFG_ENABLE bit is raised
sunhme: fix incorrect constant in sunhme_can_receive()
sunhme: add trace event for logging PCI IRQ
sun4m: set default display type to TCX
* remotes/kraxel/tags/vga-20190703-pull-request:
Add ati vgabios to INSTALL_BLOBS.
hw/i2c/bitbang_i2c: Use in-place rather than malloc'd bitbang_i2c_interface struct
virtio-gpu: check if the resource already exists in virtio_gpu_load()
virtio-gpu: fix unmap in error path
Peter Maydell [Wed, 3 Jul 2019 20:19:03 +0000 (21:19 +0100)]
Merge remote-tracking branch 'remotes/amarkovic/tags/mips-queue-jul-02-2019' into staging
MIPS queue for July 2nd, 2019
# gpg: Signature made Tue 02 Jul 2019 17:09:29 BST
# gpg: using RSA key D4972A8967F75A65
# gpg: Good signature from "Aleksandar Markovic <[email protected]>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 8526 FBF1 5DA3 811F 4A01 DD75 D497 2A89 67F7 5A65
* remotes/amarkovic/tags/mips-queue-jul-02-2019:
target/mips: Correct helper for MSA FCLASS.<W|D> instructions
target/mips: Unroll loops for MSA float max/min instructions
target/mips: Correct comments in msa_helper.c
target/mips: Correct comments in translate.c
tcg/tests: target/mips: Correct MSA test compilation and execution order
tcg/tests: target/mips: Amend MSA integer multiply tests
tcg/tests: target/mips: Amend MSA fixed point multiply tests
hw/mips: Express dependencies of the r4k platform with Kconfig
hw/mips: Express dependencies of the Jazz machine with Kconfig
hw/mips: Express dependencies of the MIPSsim machine with Kconfig
hw/mips: Explicit the semi-hosting feature is always required
tests/machine-none: Test recent MIPS cpus
* remotes/vivier2/tags/linux-user-for-4.1-pull-request:
linux-user: move QEMU_IFLA_BR_MULTI_BOOLOPT to the good function
linux-user: Handle EXCP_FPE properly for MIPS
linux-user: Introduce TARGET_HAVE_ARCH_STRUCT_FLOCK
linux-user: Fix target_flock structure for MIPS O64 ABI
linux-user: Add support for strace for statx() syscall
linux-user: Add support for translation of statx() syscall
VirtIO-RNG: Update default entropy source to `/dev/urandom`
When QEMU exposes a VirtIO-RNG device to the guest, that device needs a
source of entropy, and that source needs to be "non-blocking", like
`/dev/urandom`. However, currently QEMU defaults to the problematic
`/dev/random`, which on Linux is "blocking" (as in, it waits until
sufficient entropy is available).
Why prefer `/dev/urandom` over `/dev/random`?
---------------------------------------------
The man pages of urandom(4) and random(4) state:
"The /dev/random device is a legacy interface which dates back to a
time where the cryptographic primitives used in the implementation
of /dev/urandom were not widely trusted. It will return random
bytes only within the estimated number of bits of fresh noise in the
entropy pool, blocking if necessary. /dev/random is suitable for
applications that need high quality randomness, and can afford
indeterminate delays."
Further, the "Usage" section of the said man pages state:
"The /dev/random interface is considered a legacy interface, and
/dev/urandom is preferred and sufficient in all use cases, with the
exception of applications which require randomness during early boot
time; for these applications, getrandom(2) must be used instead,
because it will block until the entropy pool is initialized.
"If a seed file is saved across reboots as recommended below (all
major Linux distributions have done this since 2000 at least), the
output is cryptographically secure against attackers without local
root access as soon as it is reloaded in the boot sequence, and
perfectly adequate for network encryption session keys. Since reads
from /dev/random may block, users will usually want to open it in
nonblocking mode (or perform a read with timeout), and provide some
sort of user notification if the desired entropy is not immediately
available."
And refer to random(7) for a comparison of `/dev/random` and
`/dev/urandom`.
What about other OSes?
----------------------
`/dev/urandom` exists and works on OS-X, FreeBSD, DragonFlyBSD, NetBSD
and OpenBSD, which cover all the non-Linux platforms we explicitly
support, aside from Windows.
On Windows `/dev/random` doesn't work either so we don't regress.
This is actually another argument in favour of using the newly
proposed 'rng-builtin' backend by default, as that will work on
Windows.
- - -
Given the above, change the entropy source for VirtIO-RNG device to
`/dev/urandom`.
Related discussion in these[1][2] past threads.
[1] https://lists.nongnu.org/archive/html/qemu-devel/2018-06/msg08335.html
-- "RNG: Any reason QEMU doesn't default to `/dev/urandom`?"
[2] https://lists.nongnu.org/archive/html/qemu-devel/2018-09/msg02724.html
-- "[RFC] Virtio RNG: Consider changing the default entropy source to
/dev/urandom"
Makefile: Reuse all's recursion machinery for clean and install
Targets "clean" and "install" run make recursively in a for loop.
This ignores -j and -k. Target "all" depends on SUBDIR/all to recurse
into each SUBDIR. Behaves nicely with -j and -k. Put that to use for
"clean" and "install": depend on SUBDIR/clean or SUBDIR/install,
respectively, and delete the loop.
We make a few sub-directories recursively, in particular
$(TARGET_DIRS).
For goal "all", we do it the nice way: "all" has a prerequisite
subdir-T for each T in $(TARGET_DIRS), and T's recipe runs make
recursively. Behaves nicely with -j and -k.
For other goals such as "clean" and "install", the recipe runs make
recursively in a for loop. Ignores -j and -k.
The next commit will fix that for "clean" and "install". This commit
prepares the ground by renaming the targets we use for "all" to
include the goal for the sub-make. This will permit reusing them for
goals other than "all".
Targets subdir-T for T in $(TARGET_DIRS) run "make all" in T. Rename
to T/all, and declare phony.
Targets romsubdir-R for R in $(ROMS) run "make" in pc-bios/R. Default
goal is "all" for all R. Rename to pc-bios/R/all, and declare phony.
The remainder are renamed just for consistency.
Target subdir-dtc runs "make libbft/libfdt.a" in dtc. Rename to
dtc/all, and declare phony.
Target subdir-capstone runs make $(BUILD_DIR)/capstone/$(LIBCAPSTONE)
in $(SRC_PATH)/capstone. Rename to capstone/all, and declare phony.
Target subdir-slirp runs "make" in $(SRC_PATH)/slirp. Default goal is
all, which builds $(BUILD_DIR)/libslirp.a. Rename to slirp/all, and
declare phony.
Linux terminal behavior (coming from vt100 I think) is somewhat strange
when it comes to line wraps: When a character is printed to the last
char cell of a line the cursor does NOT jump to the next line but stays
where it is. The line feed happens when the next character is printed.
So the valid range for the cursor position is not 0 .. width-1 but
0 .. width, where x == width represents the state where the line is
full but the cursor didn't jump to the next line yet.
The code for the 'clear from start of line' control sequence (ESC[1K)
fails to handle this corner case correctly and may call
console_clear_xy() with x == width. That will incorrectly clear the
first char cell of the next line, or in case the cursor happens to be on
the last line overflow the cell buffer by one character (three bytes).
Add a check to the loop to fix that.
Didn't spot any other places with the same problem. But it's easy to
miss that corner case, so also allocate one extra cell as precaution, so
in case we have simliar issues lurking elsewhere it at least wouldn't be
a buffer overflow.
v2: squashed in additional checks suggested by Christophe de Dinechin.
Peter Maydell [Tue, 2 Jul 2019 16:38:44 +0000 (17:38 +0100)]
hw/i2c/bitbang_i2c: Use in-place rather than malloc'd bitbang_i2c_interface struct
Currently the bitbang_i2c_init() function allocates a
bitbang_i2c_interface struct which it returns. This is unfortunate
because it means that if the function is used from a DeviceState
init method then the memory will be leaked by an "init then delete"
cycle, as used by the qmp/hmp commands that list device properties.
Since three out of four of the uses of this function are in
device init methods, switch the function to do an in-place
initialization of a struct that can be embedded in the
device state struct of the caller.
This fixes LeakSanitizer leak warnings that have appeared in the
patchew configuration (which only tries to run the sanitizers
for the x86_64-softmmu target) now that we use the bitbang-i2c
code in an x86-64 config.
projects / qemu.git / log