Paolo Bonzini [Tue, 27 Oct 2020 08:44:18 +0000 (04:44 -0400)]
remove preconfig state
The preconfig state is only used if -incoming is not specified, which
makes the RunState state machine more tricky than it need be. However
there is already an equivalent condition which works even with -incoming,
namely qdev_hotplug. Use it instead of a separate runstate.
* remotes/ehabkost/tags/machine-next-pull-request: (25 commits)
MAINTAINERS: Update my git repository URLs
qdev: Move UUID property to qdev-properties-system.c
qdev: Make qdev_propinfo_get_uint16() static
qdev: Make error_set_from_qdev_prop_error() get Object* argument
qdev: Make check_prop_still_unset() get Object* argument
qdev: Make qdev_find_global_prop() get Object* argument
qdev: Make qdev_get_prop_ptr() get Object* arg
qdev: Make bit_prop_set() get Object* argument
qdev: Make PropertyInfo.print method get Object* argument
qdev: Don't use dev->id on set_size32() error message
sparc: Check dev->realized at sparc_set_nwindows()
qdev: Check dev->realized at set_size()
qdev: Move property code to qdev-properties.[ch]
cpu: Move cpu_common_props to hw/core/cpu.c
cs4231: Get rid of empty property array
netfilter: Use class properties
netfilter: Reorder functions
can_host: Use class properties
arm/cpu64: Register "aarch64" as class property
virt: Register "its" as class property
...
Eduardo Habkost [Fri, 11 Dec 2020 22:05:11 +0000 (17:05 -0500)]
qdev: Move UUID property to qdev-properties-system.c
Only softmmu code uses DEFINE_PROP_UUID, and it currently depends
on error_set_from_qdev_prop_error(). Move it to
qdev-properties-system.c to get out of our way when refactoring
the qdev property system.
We can eventually move it back to the core property system later,
after removing usage of error_set_from_qdev_prop_error().
Eduardo Habkost [Fri, 11 Dec 2020 22:05:03 +0000 (17:05 -0500)]
qdev: Don't use dev->id on set_size32() error message
All other qdev property error messages use "<type>.<property>"
instead of "<id>.<property>". Change set_size32() for consistency,
and to make the code not specific to TYPE_DEVICE.
Eduardo Habkost [Fri, 11 Dec 2020 22:05:02 +0000 (17:05 -0500)]
sparc: Check dev->realized at sparc_set_nwindows()
sparc_set_nwindows() is one of the very few property setters that
don't check dev->realized, and there's no reason for it to be
special. Check dev->realized like the other setters.
Eduardo Habkost [Fri, 11 Dec 2020 22:05:01 +0000 (17:05 -0500)]
qdev: Check dev->realized at set_size()
This setter is one of the very few property setters that don't
check dev->realized, and there's no reason to make size
properties different from the rest. Add the missing check.
Eduardo Habkost [Wed, 11 Nov 2020 18:38:17 +0000 (13:38 -0500)]
virt: Register "its" as class property
Class properties make QOM introspection simpler and easier, as
they don't require an object to be instantiated.
Note: "its" is currently registered conditionally, but this makes
the feature be registered unconditionally. The only side effect
is that it will be now possible to set its=on on virt-2.7 and
older.
Eduardo Habkost [Wed, 11 Nov 2020 18:38:15 +0000 (13:38 -0500)]
i386: Register feature bit properties as class properties
Class properties make QOM introspection simpler and easier, as
they don't require an object to be instantiated.
Also, the hundreds of instance properties were having an impact
on QMP commands that create temporary CPU objects. On my
machine, run time of qmp_query_cpu_definitions() changed
from ~200ms to ~16ms after applying this patch.
Volker Rümelin [Sun, 13 Dec 2020 13:05:27 +0000 (14:05 +0100)]
coreaudio: always stop audio playback on shut down
Always stop audio playback and remove the playback callback when
QEMU exits.
On shut down the function coreaudio_fini_out() destroys the
coreaudio mutex but fails to stop audio playback and to remove the
audio playback callback, because function audio_is_cleaning_up()
always returns true when called from coreaudio_fini_out(). Now
there is a time window from pthread_mutex_destroy() to program
exit where Core Audio may call the audio playback callback which
tries to lock the destroyed coreaudio mutex. This leads to the
following error.
coreaudio: Could not lock voice for audioDeviceIOProc
Reason: Invalid argument
This bug was reported on the qemu-discuss mailing list.
https://lists.nongnu.org/archive/html/qemu-discuss/2020-10/msg00018.html
Volker Rümelin [Sun, 13 Dec 2020 13:05:26 +0000 (14:05 +0100)]
coreaudio: don't start playback in init routine
Every emulated audio device has a way to enable audio playback. Don't
start playback until the guest enables the audio device to keep the
Core Audio device run state in sync with hw->enabled.
Volker Rümelin [Sun, 13 Dec 2020 13:05:25 +0000 (14:05 +0100)]
coreaudio: rename misnamed variable fake_as
While the variable once was used to fake audio settings, since
commit ed2a4a7941 "audio: proper support for float samples in
mixeng" this is no longer true. Rename the variable to obt_as.
This is the same naming scheme as in audio/sdlaudio.c
* remotes/vivier2/tags/trivial-branch-for-6.0-pull-request:
configure / meson: Move check for linux/btrfs.h to meson.build
configure / meson: Move check for sys/kcov.h to meson.build
configure / meson: Move check for sys/signal.h to meson.build
configure / meson: Move check for drm.h to meson.build
configure / meson: Move check for pty.h to meson.build
configure: Remove the obsolete check for ifaddrs.h
blockdev: Fix a memleak in drive_backup_prepare()
block/file-posix: fix a possible undefined behavior
elf2dmp/pdb: Plug memleak in pdb_init_from_file
elf2dmp/qemu_elf: Plug memleak in QEMU_Elf_init
configure: Test if $make actually exists
ads7846: moves from the hw/display folder to the hw/input folder.
CODING_STYLE.rst: Be less strict about 80 character limit
fsdev: open brace '{' following struct go on the same line
hw/pci-host/pam: Replace magic number by PAM_REGIONS_COUNT definition
hw/xen: Don't use '#' flag of printf format
MAINTAINERS: update my email address
qemu-options.hx: Fix minor issues in icount documentation
target/i386: tracing: format length values as hex
Peter Maydell [Mon, 14 Dec 2020 16:31:15 +0000 (16:31 +0000)]
Merge remote-tracking branch 'remotes/dg-gitlab/tags/ppc-for-6.0-20201214' into staging
ppc patch queue 2020-12-14
Here's my first pull request for qemu-6.0, with a bunch of things
queued over the freeze. Highlights are:
* A bunch of cleanups to hotplug error paths from Greg Kurz
* A number of TCG fixes from new contributor Giuseppe Musacchio
* Added Greg Kurz as co-maintainer
* Assorted other bugfixes and cleanups
This supersedes ppc-for-6.0-20201211, the only change are some patch
authors to better match qemu conventions.
* remotes/dg-gitlab/tags/ppc-for-6.0-20201214: (30 commits)
spapr.c: set a 'kvm-type' default value instead of relying on NULL
spapr: Pass sPAPR machine state to some RTAS events handling functions
spapr: Don't use qdev_get_machine() in spapr_msi_write()
spapr: Pass sPAPR machine state down to spapr_pci_switch_vga()
target/ppc: Introduce an mmu_is_64bit() helper
ppc/translate: Use POWERPC_MMU_64 to detect 64-bit MMU models
ppc/e500: Free irqs array to avoid memleak
MAINTAINERS: Add Greg Kurz as co-maintainer for ppc
hw/ppc: Do not re-read the clock on pre_save if doing savevm
target/ppc: Remove "compat" property of server class POWER CPUs
spapr: spapr_drc_attach() cannot fail
spapr: Simplify error path of spapr_core_plug()
spapr: Abort if ppc_set_compat() fails for hot-plugged CPUs
spapr: Fix pre-2.10 dummy ICP hack
xive: Add trace events
hw/ppc/spapr_tpm_proxy: Fix hexadecimal format string specifier
ppc/translate: Rewrite gen_lxvdsx to use gvec primitives
ppc/translate: Raise exceptions after setting the cc
ppc/translate: Delay NaN checking after comparison
ppc/translate: Turn the helper macros into functions
...
Peter Maydell [Mon, 14 Dec 2020 13:37:02 +0000 (13:37 +0000)]
tests/tcg/multiarch/Makefile.target: Disable run-gdbstub-sha1 test
Disable the run-gdbstub-sha1 test: it provokes an internal error
assertion failure in Ubuntu gdb 8.1.1-0ubuntu1 (Ubuntu gdb
8.1-0ubuntu3.2 also has this assert but we were previously skipping
this test because it doesn't support connection over local domain
sockets) :
timeout 60 /home/petmay01/linaro/qemu-for-merges/tests/guest-debug/run-test.py --gdb /usr/bin/gdb-multiar
/build/gdb-veKdC1/gdb-8.1.1/gdb/regcache.c:122: internal-error: void* init_regcache_descr(gdbarch*): Asser
A problem internal to GDB has been detected,
further debugging may prove unreliable.
This is a bug, please report it. For instructions, see:
<http://www.gnu.org/software/gdb/bugs/>.
Aborted (core dumped)
/home/petmay01/linaro/qemu-for-merges/tests/tcg/multiarch/Makefile.target:51: recipe for target 'run-gdbst
spapr.c: set a 'kvm-type' default value instead of relying on NULL
spapr_kvm_type() is considering 'vm_type=NULL' as a valid input, where
the function returns 0. This is relying on the current QEMU machine
options handling logic, where the absence of the 'kvm-type' option
will be reflected as 'vm_type=NULL' in this function.
This is not robust, and will break if QEMU options code decides to propagate
something else in the case mentioned above (e.g. an empty string instead
of NULL).
Let's avoid this entirely by setting a non-NULL default value in case of
no user input for 'kvm-type'. spapr_kvm_type() was changed to handle 3 fixed
values of kvm-type: "auto", "hv", and "pr", with "auto" being the default
if no kvm-type was set by the user. This allows us to always be predictable
regardless of any enhancements/changes made in QEMU options mechanics.
While we're at it, let's also document in 'kvm-type' description the
already existing default mode, now named 'auto'. The information provided
about it is based on how the pseries kernel handles the KVM_CREATE_VM
ioctl(), where the default value '0' makes the kernel choose an available
KVM module to use, giving precedence to kvm_hv. This logic is described in
the kernel source file arch/powerpc/kvm/powerpc.c, function kvm_arch_init_vm().
Greg Kurz [Wed, 9 Dec 2020 17:00:51 +0000 (18:00 +0100)]
spapr: Pass sPAPR machine state to some RTAS events handling functions
Some functions in hw/ppc/spapr_events.c get a pointer to the machine
state using qdev_get_machine(). Convert them to get it from their
caller when possible.
ppc/translate: Use POWERPC_MMU_64 to detect 64-bit MMU models
The ppc_tr_init_disas_context() function currently checks whether the
MMU is 64-bit by ANDing its model type with POWERPC_MMU_64B. This is
wrong : POWERPC_MMU_64B isn't a mask, it is the generic MMU model for
pre-PowerISA-2.03 64-bit CPUs (ie. PowerPC 970 in QEMU).
Use POWERPC_MMU_64 instead of POWERPC_MMU_64B. This should fix a
potential bug with some 32-bit CPUs for which 'need_access_type'
was mis-computed because (POWERPC_MMU_32B & POWERPC_MMU_64B)
happens to be equal to 1. The end result being a crash in
ppc_hash32_direct_store() because the access type isn't set:
cpu_abort(cs, "ERROR: instruction should not need "
"address translation\n");
This doesn't change anything for 'lazy_tlb_flush' since POWERPC_MMU_32B
is checked first.
Fixes: 5f2a6254522b ("ppc: Don't set access_type on all load/stores on hash64") Signed-off-by: Stephane Duverger <[email protected]>
[groug: - extended patch to address another misuse of POWERPC_MMU_64B
- updated title and changelog accordingly] Signed-off-by: Greg Kurz <[email protected]>
Message-Id: <20201209173536.1437351[email protected]> Signed-off-by: David Gibson <[email protected]>
Gan Qixin [Fri, 4 Dec 2020 07:58:22 +0000 (15:58 +0800)]
ppc/e500: Free irqs array to avoid memleak
When running qom-test, a memory leak occurred in the ppce500_init function,
this patch free irqs array to fix it.
ASAN shows memory leak stack:
Direct leak of 40 byte(s) in 1 object(s) allocated from:
#0 0xfffc5ceee1f0 in __interceptor_calloc (/lib64/libasan.so.5+0xee1f0)
#1 0xfffc5c806800 in g_malloc0 (/lib64/libglib-2.0.so.0+0x56800)
#2 0xaaacf9999244 in ppce500_init qemu/hw/ppc/e500.c:859
#3 0xaaacf97434e8 in machine_run_board_init qemu/hw/core/machine.c:1134
#4 0xaaacf9c9475c in qemu_init qemu/softmmu/vl.c:4369
#5 0xaaacf94785a0 in main qemu/softmmu/main.c:49
David Gibson [Thu, 26 Nov 2020 04:09:16 +0000 (15:09 +1100)]
MAINTAINERS: Add Greg Kurz as co-maintainer for ppc
Greg has agreed to be co-maintainer of the ppc target and machines.
This should avoid repeats of the problem we had in qemu-5.2 where a
last minute fix was needed while I was on holiday.
Greg Kurz [Wed, 2 Dec 2020 17:28:26 +0000 (18:28 +0100)]
hw/ppc: Do not re-read the clock on pre_save if doing savevm
A guest with enough RAM, eg. 128G, is likely to detect savevm downtime
and to complain about stalled CPUs. This happens because we re-read
the timebase just before migrating it and we thus don't account for
all the time between VM stop and pre-save.
A very similar situation was already addressed for live migration of
paused guests (commit d14f33976282). Extend the logic to do the same
with savevm.
Greg Kurz [Tue, 1 Dec 2020 13:11:03 +0000 (14:11 +0100)]
target/ppc: Remove "compat" property of server class POWER CPUs
This property has been deprecated since QEMU 5.0 by commit 22062e54bb68.
We only kept a legacy hack that internally converts "compat" into the
official "max-cpu-compat" property of the pseries machine type.
According to our deprecation policy, we could have removed it for QEMU 5.2
already. Do it now ; since ppc_cpu_parse_featurestr() now just calls the
generic parent_parse_features handler, drop it as well.
Users are supposed to use the "max-cpu-compat" property of the pseries
machine type instead.
Greg Kurz [Tue, 1 Dec 2020 11:37:28 +0000 (12:37 +0100)]
spapr: spapr_drc_attach() cannot fail
All users are passing &error_abort already. Document the fact
that spapr_drc_attach() should only be passed a free DRC, which
is supposedly the case if appropriate checking is done earlier.
Greg Kurz [Tue, 1 Dec 2020 11:37:27 +0000 (12:37 +0100)]
spapr: Simplify error path of spapr_core_plug()
spapr_core_pre_plug() already guarantees that the slot for the given core
ID is available. It is thus safe to assume that spapr_find_cpu_slot()
returns a slot during plug. Turn the error path into an assertion.
It is also safe to assume that no device is attached to the corresponding
DRC and that spapr_drc_attach() shouldn't fail.
Pass &error_abort to spapr_drc_attach() and simplify error handling.
Greg Kurz [Tue, 1 Dec 2020 11:37:26 +0000 (12:37 +0100)]
spapr: Abort if ppc_set_compat() fails for hot-plugged CPUs
When a CPU is hot-plugged, we set its compat mode to match the boot
CPU, which was either set by machine reset or by CAS. This is currently
handled in the plug handler after the core got realized. Potential errors
of ppc_set_compat() are propagated to the hot-plug logic.
Handling errors this late in the hot-plug sequence is generally frown
upon. Ideally, we should do sanity checks in a pre-plug handler and pass
&error_abort to ppc_set_compat() in the plug handler.
We can filter out some error cases of ppc_set_compat() by calling
ppc_check_compat() at pre-plug. But ppc_set_compat() also sets the
compat register in KVM, and KVM doesn't provide any API that would
allow to check valid compat mode settings beforehand.
However, at this point we know that the compat mode was already
successfully set for the boot CPU. Since this all boils down to
setting a register with the very same value that was valid
for the boot CPU, it should definitely not fail for hot-plugged
CPUS.
Greg Kurz [Tue, 1 Dec 2020 11:37:25 +0000 (12:37 +0100)]
spapr: Fix pre-2.10 dummy ICP hack
This hack registers dummy VMState entries of ICPs in order to
support migration of old pseries machine types that used to
create all smp.max_cpus possible ICPs at machine init.
Part of the work is to unregister the dummy entries when plugging
an actual vCPU core, and to register them back when unplugging the
core. The code that unregisters the dummy ICPs in spapr_core_plug()
is misplaced: if ppc_set_compat() fails afterwards, the hotplug
operation will be cancelled and the dummy ICPs won't be registered
back since the unplug handler isn't called.
Unregister the dummy ICPs at the end of spapr_core_plug().
ppc/translate: Rewrite gen_lxvdsx to use gvec primitives
Make the implementation match the lxvwsx one.
The code is now shorter smaller and potentially faster as the
translation will use the host SIMD capabilities if available.
According to the PowerISA v3.1 reference, Table 68 "Actions for xscmpudp
- Part 1: Compare Unordered", whenever one of the two operands is a NaN
the SO bit is set while the other three bits are cleared.
Apply the same change to xscmpuqp.
The respective ordered counterparts are unaffected.
Chen Qun [Mon, 16 Nov 2020 02:48:09 +0000 (10:48 +0800)]
ppc: Add a missing break for PPC6xx_INPUT_TBEN
When using -Wimplicit-fallthrough in our CFLAGS, the compiler showed warning:
hw/ppc/ppc.c: In function ‘ppc6xx_set_irq’:
hw/ppc/ppc.c:118:16: warning: this statement may fall through [-Wimplicit-fallthrough=]
118 | if (level) {
| ^
hw/ppc/ppc.c:123:9: note: here
123 | case PPC6xx_INPUT_INT:
| ^~~~
According to the discussion, a break statement needs to be added here.
Chen Qun [Mon, 16 Nov 2020 02:48:10 +0000 (10:48 +0800)]
target/ppc: replaced the TODO with LOG_UNIMP and add break for silence warnings
When using -Wimplicit-fallthrough in our CFLAGS, the compiler showed warning:
target/ppc/mmu_helper.c: In function ‘dump_mmu’:
target/ppc/mmu_helper.c:1351:12: warning: this statement may fall through [-Wimplicit-fallthrough=]
1351 | if (ppc64_v3_radix(env_archcpu(env))) {
| ^
target/ppc/mmu_helper.c:1358:5: note: here
1358 | default:
| ^~~~~~~
Use "qemu_log_mask(LOG_UNIMP**)" instead of the TODO comment.
And add the break statement to fix it.
Greg Kurz [Fri, 20 Nov 2020 23:42:01 +0000 (00:42 +0100)]
spapr: Do NVDIMM/PC-DIMM device hotplug sanity checks at pre-plug only
Pre-plug of a memory device, be it an NVDIMM or a PC-DIMM, ensures
that the memory slot is available and that addresses don't overlap
with existing memory regions. The corresponding DRCs in the LMB
and PMEM namespaces are thus necessarily attachable at plug time.
Pass &error_abort to spapr_drc_attach() in spapr_add_lmbs() and
spapr_add_nvdimm(). This allows to greatly simplify error handling
on the plug path.
Greg Kurz [Fri, 20 Nov 2020 23:42:00 +0000 (00:42 +0100)]
spapr: Do PCI device hotplug sanity checks at pre-plug only
The PHB acts as the hotplug handler for PCI devices. It does some
sanity checks on DR enablement, PCI bridge chassis numbers and
multifunction. These checks are currently performed at plug time,
but they would best sit in a pre-plug handler in order to error
out as early as possible.
Create a spapr_pci_pre_plug() handler and move all the checking
there. Add a check that the associated DRC doesn't already have
an attached device. This is equivalent to the slot availability
check performed by do_pci_register_device() upon realization of
the PCI device.
This allows to pass &error_abort to spapr_drc_attach() and to end
up with a plug handler that doesn't need to report errors anymore.
Greg Kurz [Fri, 20 Nov 2020 17:46:39 +0000 (18:46 +0100)]
spapr/xive: Turn some sanity checks into assertions
The sPAPR XIVE device is created by the machine in spapr_irq_init().
The latter overrides any value provided by the user with -global for
the "nr-irqs" and "nr-ends" properties with strictly positive values.
It seems reasonable to assume these properties should never be 0,
which wouldn't make much sense by the way.
Thomas Huth [Wed, 18 Nov 2020 17:10:52 +0000 (18:10 +0100)]
configure / meson: Move check for linux/btrfs.h to meson.build
This check can be done in a much shorter way in meson.build. And while
we're at it, rename the #define to HAVE_BTRFS_H to match the other
HAVE_someheader_H symbols that we already have.
Thomas Huth [Wed, 18 Nov 2020 17:10:51 +0000 (18:10 +0100)]
configure / meson: Move check for sys/kcov.h to meson.build
This check can be done in a much shorter way in meson.build. And while
we're at it, rename the #define to HAVE_SYS_KCOV_H to match the other
HAVE_someheader_H symbols that we already have.
Thomas Huth [Wed, 18 Nov 2020 17:10:47 +0000 (18:10 +0100)]
configure: Remove the obsolete check for ifaddrs.h
The code that used HAVE_IFADDRS_H has been removed in commit 0a27af918b ("io: use bind() to check for IPv4/6 availability"),
so we don't need this check in the configure script anymore.
Roman Bolshakov [Tue, 25 Aug 2020 20:27:55 +0000 (23:27 +0300)]
configure: Test if $make actually exists
configure doesn't detect if $make is installed on the build host.
This is also helpful for hosts where an alias for make is used, i.e.
configure would fail if gmake is not present on macOS.
hw/mips/malta: Rewrite CP0_MVPConf0 access using deposit()
PTC field has 8 bits, PVPE has 4. We plan to use the
"hw/registerfields.h" API with MIPS CPU definitions
(target/mips/cpu.h). Meanwhile we use magic 8 and 4.
Introduce cpu_supports_isa() which takes a CPUMIPSState
argument, more useful at runtime when the CPU is created
(no need to call the extensive object_class_by_name()).
projects / qemu.git / log