Reduce the amount of preprocessor obfuscation by expanding
the text of each of the functions generated. The result is
only slightly smaller than the original.
target/ppc: Use cpu_*_mmuidx_ra instead of MMU_MODE*_SUFFIX
There are only two uses. Within dcbz_common, the local variable
mmu_idx already contains the epid computation, and we can avoid
repeating it for the store. Within helper_icbiep, the usage is
trivially expanded using PPC_TLB_EPID_LOAD.
target/s390x: Use cpu_*_mmuidx_ra instead of MMU_MODE*_SUFFIX
The generated functions aside from *_real are unused.
The *_real functions have a couple of users in mem_helper.c;
use *_mmuidx_ra instead, with MMU_REAL_IDX.
target/mips: Use cpu_*_mmuidx_ra instead of MMU_MODE*_SUFFIX
The separate suffixed functions were used to construct
some do_##insn function switched on mmu_idx. The interface
is exactly identical to the *_mmuidx_ra functions. Replace
them directly and remove the constructions.
cputlb: Expand cpu_ldst_useronly_template.h in user-exec.c
With the tracing hooks, the inline functions are no longer
so simple. Reduce the amount of preprocessor obfuscation
by expanding the text of each of the functions generated.
cputlb: Rename helper_ret_ld*_cmmu to cpu_ld*_code
There are no uses of the *_cmmu names other than the bare wrapping
within the *_code inlines. Therefore rename the functions so we
can drop the inlines.
Use abi_ptr instead of target_ulong in preparation for user-only;
the two types are identical for softmmu.
cputlb: Move body of cpu_ldst_template.h out of line
With the tracing hooks, the inline functions are no longer
so simple. Once out-of-line, the current tlb_entry lookup
is redundant with the one in the main load/store_helper.
This also begins the introduction of a new target facing
interface, with suffix *_mmuidx_ra. This is not yet
official because the interface is not done for user-only.
Use abi_ptr instead of target_ulong in preparation for
user-only; the two types are identical for softmmu.
What remains in cpu_ldst_template.h are the expansions
for _code, _data, and MMU_MODE<N>_SUFFIX.
It is easy for the atomic helpers to use trace_mem_build_info
directly, without resorting to symbol pasting. For this usage,
we cannot use trace_mem_get_info, because the MemOp does not
support 16-byte accesses.
cputlb: Use trace_mem_get_info instead of trace_mem_build_info
In the cpu_ldst templates, we already require a MemOp, and it
is cleaner and clearer to pass that instead of 3 separate
arguments describing the memory operation.
Nikola Pavlica [Wed, 8 Jan 2020 12:13:42 +0000 (13:13 +0100)]
display/gtk: get proper refreshrate
Because some VMs in QEMU can get GPU virtualization (using technologies
such as iGVT-g, as mentioned previously), they could produce a video
output that had a higher display refresh rate than of what the GTK
display was displaying. (fxp. Playing a video game inside of a Windows
VM at 60 Hz, while the output stood locked at 33 Hz because of defaults
set in include/ui/console.h)
Since QEMU does indeed have internal systems for determining frame
times as defined in ui/console.c.
The code checks for a variable called update_interval that it later
uses for time calculation. This variable, however, isn't defined
anywhere in ui/gtk.c and instead ui/console.c just sets it to
GUI_REFRESH_INTERVAL_DEFAULT which is 30
update_interval represents the number of milliseconds per display
refresh, and by doing some math we get that 1000/30 = 33.33... Hz
This creates the mentioned problem and what this patch does is that it
checks for the display refresh rate reported by GTK itself (we can take
this as a safe value) and just converts it back to a number of
milliseconds per display refresh.
Thomas Huth [Wed, 8 Jan 2020 14:47:02 +0000 (15:47 +0100)]
ui: Print available display backends with '-display help'
We already print availabled devices with "-device help", or available
backends with "-netdev help" or "-chardev help". Let's provide a way
for the users to query the available display backends, too.
Stefan Hajnoczi [Tue, 7 Jan 2020 11:24:38 +0000 (11:24 +0000)]
trace: update qemu-trace-stap to Python 3
qemu-trace-stap does not support Python 3 yet:
$ scripts/qemu-trace-stap list path/to/qemu-system-x86_64
Traceback (most recent call last):
File "scripts/qemu-trace-stap", line 175, in <module>
main()
File "scripts/qemu-trace-stap", line 171, in main
args.func(args)
File "scripts/qemu-trace-stap", line 118, in cmd_list
print_probes(args.verbose, "*")
File "scripts/qemu-trace-stap", line 114, in print_probes
if line.startswith(prefix):
TypeError: startswith first arg must be bytes or a tuple of bytes, not str
Now that QEMU requires Python 3.5 or later we can switch to pure Python
3. Use Popen()'s universal_newlines=True argument to treat stdout as
text instead of binary.
Wangyong [Tue, 7 Jan 2020 06:01:01 +0000 (06:01 +0000)]
linux-aio: increasing MAX_EVENTS to a larger hardcoded value
Since commit 6040aedddb5f474a9c2304b6a432a652d82b3d3c "virtio-blk:
make queue size configurable",if the user set the queue size to
more than 128 ,it will not take effect. That's because linux aio's
maximum outstanding requests at a time is always less than or equal
to 128.
This patch simply increase MAX_EVENTS to a larger hardcoded value of
1024 as a shortterm fix.
* remotes/kraxel/tags/usb-20200113-pull-request:
xhci: recheck slot status
xhci: Fix memory leak in xhci_kick_epctx when poweroff GuestOS
usbredir: Prevent recursion in usbredir_write
usb-redir: remove 'remote wakeup' flag from configuration descriptor
usb-host: remove 'remote wakeup' flag from configuration descriptor
* remotes/huth-gitlab/tags/pull-request-2020-01-12:
docs: build an index page for the HTML docs
tests/libqos: Move the libqos files under tests/qtest/
tests/Makefile: Move qtest-related settings to a separate Makefile.include
test: Move qtests to a separate directory
tests/Makefile: Separate unit test dependencies from qtest dependencies
tests/Makefile: Remove 'tests/' and '$(EXESUF)' from the check-qtest variables
tests/ptimer: Remove unnecessary inclusion of libqtest.h
tests/Makefile: test-char does not need libqtest
Gerd Hoffmann [Tue, 7 Jan 2020 08:36:06 +0000 (09:36 +0100)]
xhci: recheck slot status
Factor out slot status check into a helper function. Add an additional
check after completing transfers. This is needed in case a guest
queues multiple transfers in a row and a device unplug happens while
qemu processes them.
I've got a case where usbredir_write manages to call back into itself
via spice; this patch causes the recursion to fail (0 bytes) the write;
this seems to avoid the deadlock I was previously seeing.
I can't say I fully understand the interaction of usbredir and spice;
but there are a few similar guards in spice and usbredir
to catch other cases especially onces also related to spice_server_char_device_wakeup
This case seems to be triggered by repeated migration+repeated
reconnection of the viewer; but my debugging suggests the migration
finished before this hits.
The backtrace of the hang looks like:
reds_handle_ticket
reds_handle_other_links
reds_channel_do_link
red_channel_connect
spicevmc_connect
usbredir_create_parser
usbredirparser_do_write
usbredir_write
qemu_chr_fe_write
qemu_chr_write
qemu_chr_write_buffer
spice_chr_write
spice_server_char_device_wakeup
red_char_device_wakeup
red_char_device_write_to_device
vmc_write
usbredirparser_do_write
usbredir_write
qemu_chr_fe_write
qemu_chr_write
qemu_chr_write_buffer
qemu_mutex_lock_impl
and we fail as we lang through qemu_chr_write_buffer's lock
twice.
usb-redir: remove 'remote wakeup' flag from configuration descriptor
If the redirected device has this capability, Windows guest may
place the device into D2 and expect it to wake when the device
becomes active, but this will never happen. For example, when
internal Bluetooth adapter is redirected, keyboards and mice
connected to it do not work. Current commit removes this
capability (starting from machine 5.0)
Set 'usb-redir.suppress-remote-wake' property to 'off' to keep
'remote wake' as is or to 'on' to remove 'remote wake' on
4.2 or earlier.
usb-host: remove 'remote wakeup' flag from configuration descriptor
If the redirected device has this capability, Windows guest may
place the device into D2 and expect it to wake when the device
becomes active, but this will never happen. For example, when
internal Bluetooth adapter is redirected, keyboards and mice
connected to it do not work. Current commit removes this
capability (starting from machine 5.0)
Set 'usb-host.suppress-remote-wake' property to 'off' to keep
'remote wake' as is or to 'on' to remove 'remote wake' on
4.2 or earlier.
Stefan Hajnoczi [Thu, 19 Dec 2019 13:56:20 +0000 (13:56 +0000)]
docs: build an index page for the HTML docs
There is no index.html start page for the QEMU HTML documentation. An
index page is needed so that documentation can be browsed easily on the
web.
This patch adds an index.html.in template file where the QEMU version
number is expanded. It is written in HTML instead of using the existing
sphinx (rST) and texi documentation generators because they are
heavyweight and would make this harder.
Thomas Huth [Mon, 9 Sep 2019 12:41:27 +0000 (14:41 +0200)]
tests/Makefile: Move qtest-related settings to a separate Makefile.include
tests/Makefile.include is pretty much overcrowded. Now that we have a
dedicated folder for the qtests, let's move the related settings
to a Makefile.include file in that directory instead.
Thomas Huth [Mon, 9 Sep 2019 10:04:01 +0000 (12:04 +0200)]
test: Move qtests to a separate directory
The tests directory itself is pretty overcrowded, and it's hard to
see which test belongs to which test subsystem (unit, qtest, ...).
Let's move the qtests to a separate folder for more clarity.
* remotes/bonzini/tags/for-upstream: (38 commits)
chardev: Use QEMUChrEvent enum in IOEventHandler typedef
chardev: use QEMUChrEvent instead of int
chardev/char: Explicit we ignore some QEMUChrEvent in IOEventHandler
monitor/hmp: Explicit we ignore a QEMUChrEvent in IOEventHandler
monitor/qmp: Explicit we ignore few QEMUChrEvent in IOEventHandler
virtio-console: Explicit we ignore some QEMUChrEvent in IOEventHandler
vhost-user-blk: Explicit we ignore few QEMUChrEvent in IOEventHandler
vhost-user-net: Explicit we ignore few QEMUChrEvent in IOEventHandler
vhost-user-crypto: Explicit we ignore some QEMUChrEvent in IOEventHandler
ccid-card-passthru: Explicit we ignore QEMUChrEvent in IOEventHandler
hw/usb/redirect: Explicit we ignore few QEMUChrEvent in IOEventHandler
hw/usb/dev-serial: Explicit we ignore few QEMUChrEvent in IOEventHandler
hw/char/terminal3270: Explicit ignored QEMUChrEvent in IOEventHandler
hw/ipmi: Explicit we ignore some QEMUChrEvent in IOEventHandler
hw/ipmi: Remove unnecessary declarations
target/i386: Add missed features to Cooperlake CPU model
target/i386: Add new bit definitions of MSR_IA32_ARCH_CAPABILITIES
target/i386: Fix handling of k_gs_base register in 32-bit mode in gdbstub
hw/rtc/mc146818: Add missing dependency on ISA Bus
hw/nvram/Kconfig: Restrict CHRP NVRAM to machines using OpenBIOS or SLOF
...
Peter Maydell [Fri, 10 Jan 2020 16:15:04 +0000 (16:15 +0000)]
Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-5.0-20200108' into staging
ppc patch queue 2020-01-08
Here's another pull request for qemu-5.0 of ppc related changes.
Highlights are:
* First parts of support for POWER Secure VMs
* Rework to clean up how we pass context information to the various
components of the pnv machine (reduces usage of qdev_get_machine())
* Assorted cleanups and bugfixes
* remotes/dgibson/tags/ppc-for-5.0-20200108: (26 commits)
ppc/pnv: fix check on return value of blk_getlength()
ppc/pnv: check return value of blk_pwrite()
pnv/psi: Consolidate some duplicated code in pnv_psi_realize()
pnv/psi: Add device reset hook
pnv/xive: Deduce the PnvXive pointer from XiveTCTX::xptr
spapr/xive: Deduce the SpaprXive pointer from XiveTCTX::xptr
xive: Add a "presenter" link property to the TCTX object
ppc/pnv: Add a "pnor" const link property to the BMC internal simulator
ppc/pnv: Add an "nr-threads" property to the base chip class
xive: Use the XIVE fabric link under the XIVE router
spapr, pnv, xive: Add a "xive-fabric" link to the XIVE router
pnv/xive: Use device_class_set_parent_realize()
ppc/pnv: Introduce a "xics" property under the POWER8 chip
ppc/pnv: Introduce a "xics" property alias under the PSI model
spapr/xive: remove redundant check in spapr_match_nvt()
ppc/pnv: Drop "num-chips" machine property
ppc440_bamboo.c: remove label from bamboo_load_device_tree()
spapr.c: remove 'out' label in spapr_dt_cas_updates()
ppc/spapr: Don't call KVM_SVM_OFF ioctl on TCG
spapr/xive: Use device_class_set_parent_realize()
...
Peter Maydell [Fri, 10 Jan 2020 13:19:34 +0000 (13:19 +0000)]
Merge remote-tracking branch 'remotes/stsquad/tags/pull-testing-and-semihosting-090120-2' into staging
Testing fixes and semiconsole support:
- build fix (missing x86-iommu stubs)
- python fixes for freebsd and OSX
- nicer reporting of acceptance failures
- fix build nesting of fp-test (breaks bsds)
- semihosting clean-ups
- support for blocking semihosting console
# gpg: Signature made Thu 09 Jan 2020 11:42:51 GMT
# 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-and-semihosting-090120-2:
tests/tcg: add user version of dumb-as-bricks semiconsole test
tests/tcg: extract __semi_call into a header and expand
tests/tcg: add a dumb-as-bricks semihosting console test
semihosting: add qemu_semihosting_console_inc for SYS_READC
target/arm: only update pc after semihosting completes
target/arm: remove unused EXCP_SEMIHOST leg
testing: don't nest build for fp-test
travis.yml: install homebrew python for OS X
travis.yml: duplicate before_script for MacOSX
travis.yml: Detach build and test steps
travis.yml: avocado: Print logs of non-pass tests only
freebsd: use python37
tests/vm: update openbsd to release 6.6
hw/i386/x86-iommu: Add missing stubs
Alex Bennée [Wed, 18 Dec 2019 14:28:36 +0000 (14:28 +0000)]
tests/tcg: add a dumb-as-bricks semihosting console test
We don't run this during check-tcg as we would need to check stuff is
echoed back. However we can still build the binary so people can test
it manually.
Keith Packard [Mon, 4 Nov 2019 20:42:30 +0000 (12:42 -0800)]
semihosting: add qemu_semihosting_console_inc for SYS_READC
Provides a blocking call to read a character from the console using
semihosting.chardev, if specified. This takes some careful command
line options to use stdio successfully as the serial ports, monitor
and semihost all want to use stdio. Here's a sample set of command
line options which share stdio between semihost, monitor and serial
ports:
Alex Bennée [Tue, 17 Dec 2019 15:08:57 +0000 (15:08 +0000)]
target/arm: only update pc after semihosting completes
Before we introduce blocking semihosting calls we need to ensure we
can restart the system on semi hosting exception. To be able to do
this the EXCP_SEMIHOST operation should be idempotent until it finally
completes. Practically this means ensureing we only update the pc
after the semihosting call has completed.
Alex Bennée [Tue, 7 Jan 2020 17:55:52 +0000 (17:55 +0000)]
testing: don't nest build for fp-test
Re-calling the main make is counter-productive and really messes up
with parallel builds. Just ensure we have built the pre-requisites
before we build the fp-test bits. If the user builds manually just
complain if the parent build hasn't got the bits we need.
Alex Bennée [Tue, 7 Jan 2020 13:38:28 +0000 (13:38 +0000)]
travis.yml: install homebrew python for OS X
Our python3 requirements now outstrip those of the build. While we are
at it we can move more of the special casing for Mac into the one
build we have.
Currently build and test commands are a single step in a
Travis's `script` block. In order to see the output
of the tests one needs to scroll down the log to find where
the build messages ended and the limit is not clear. If
they were in different steps then Travis would print the
result build command, which can be easily grep'ed.
So this change is made to detach those commands
to ease the visualization of the output.
Note that all steps on the `script` block is executed regardless
if one previous has failed. To overcome it, let's save the
return code of the build then check whether succeed or failed on
the test step.
travis.yml: avocado: Print logs of non-pass tests only
The acceptance tests build on Travis is configured to print
the entire Avocado's job log in case any test fail. Usually one is
interested on failed tests only though. So this change the Travis
configuration in order to show the log of tests which status is
different from 'PASS' and 'SKIP' only. Note that 'CANCEL'-ed tests
will have the log printed too because it can help to debug some
condition on CI environment which is not being fulfilled.
Gerd Hoffmann [Mon, 6 Jan 2020 12:37:46 +0000 (13:37 +0100)]
freebsd: use python37
FreeBSD seems to use python37 by default now, which breaks the build
script. Add python to the package list, to explicitly pick the version,
and also adapt the configure command line.
In commit 6c730e4af9 we introduced a stub to build the MicroVM
machine without Intel IOMMU. This stub is incomplete for the
other PC machines. Add the missing stubs.
Chen Qun [Wed, 8 Jan 2020 11:42:07 +0000 (19:42 +0800)]
vl: fix memory leak in configure_accelerators
The accel_list forgot to free, the asan output:
Direct leak of 16 byte(s) in 1 object(s) allocated from:
#0 0xffff919331cb in __interceptor_malloc (/lib64/libasan.so.4+0xd31cb)
#1 0xffff913f7163 in g_malloc (/lib64/libglib-2.0.so.0+0x57163)
#2 0xffff91413d9b in g_strsplit (/lib64/libglib-2.0.so.0+0x73d9b)
#3 0xaaab42fb58e7 in configure_accelerators /qemu/vl.c:2777
#4 0xaaab42fb58e7 in main /qemu/vl.c:4121
#5 0xffff8f9b0b9f in __libc_start_main (/lib64/libc.so.6+0x20b9f)
#6 0xaaab42fc1dab (/qemu/build/aarch64-softmmu/qemu-system-aarch64+0x8b1dab)
Indirect leak of 4 byte(s) in 1 object(s) allocated from:
#0 0xffff919331cb in __interceptor_malloc (/lib64/libasan.so.4+0xd31cb)
#1 0xffff913f7163 in g_malloc (/lib64/libglib-2.0.so.0+0x57163)
#2 0xffff9141243b in g_strdup (/lib64/libglib-2.0.so.0+0x7243b)
#3 0xffff91413e6f in g_strsplit (/lib64/libglib-2.0.so.0+0x73e6f)
#4 0xaaab42fb58e7 in configure_accelerators /qemu/vl.c:2777
#5 0xaaab42fb58e7 in main /qemu/vl.c:4121
#6 0xffff8f9b0b9f in __libc_start_main (/lib64/libc.so.6+0x20b9f)
#7 0xaaab42fc1dab (/qemu/build/aarch64-softmmu/qemu-system-aarch64+0x8b1dab)
Fixes:
target/arm/translate-a64.c: In function 'disas_crypto_three_reg_sha512':
target/arm/translate-a64.c:13625:9: error: 'genfn' may be used uninitialized in this function [-Werror=maybe-uninitialized]
genfn(tcg_rd_ptr, tcg_rn_ptr, tcg_rm_ptr);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
qemu/target/arm/translate-a64.c:13609:8: error: 'feature' may be used uninitialized in this function [-Werror=maybe-uninitialized]
if (!feature) {
Pan Nengyuan [Wed, 8 Jan 2020 02:51:32 +0000 (10:51 +0800)]
nbd: fix uninitialized variable warning
Fixes:
/mnt/sdb/qemu/nbd/server.c: In function 'nbd_handle_request':
/mnt/sdb/qemu/nbd/server.c:2313:9: error: 'ret' may be used uninitialized in this function [-Werror=maybe-uninitialized]
int ret;
chardev: Use QEMUChrEvent enum in IOEventHandler typedef
The Chardev events are listed in the QEMUChrEvent enum.
By using the enum in the IOEventHandler typedef we:
- make the IOEventHandler type more explicit (this handler
process out-of-band information, while the IOReadHandler
is in-band),
- help static code analyzers.
This patch was produced with the following spatch script:
Paolo Bonzini [Wed, 8 Jan 2020 10:14:29 +0000 (11:14 +0100)]
chardev: use QEMUChrEvent instead of int
This uses the QEMUChrEvent enum everywhere except in IOEventHandler.
The IOEventHandler change needs to happen at once for all front ends and
is done with Coccinelle in the next patch.
(Extracted from a patch by Philippe Mathieu-Daudé).
ppc/pnv: fix check on return value of blk_getlength()
blk_getlength() returns an int64_t but the result is stored in a
uint32_t. Errors (negative values) won't be caught by the check in
pnv_pnor_realize() and blk_blockalign() will allocate a very large
buffer in such cases.
Greg Kurz [Tue, 7 Jan 2020 16:32:46 +0000 (17:32 +0100)]
pnv/psi: Consolidate some duplicated code in pnv_psi_realize()
The proper way to do that would be to use device_class_set_parent_realize(),
but defining a Pnv8PsiClass and a Pnv9PsiClass types with a parent_realize
pointer adds a fair amount of code. Calling pnv_psi_realize() explicitely
is fine for now.
This should probably be achieved with a device realize hook in the
PSI base class and device_class_set_parent_realize() in the children
classes.
Greg Kurz [Tue, 7 Jan 2020 16:32:40 +0000 (17:32 +0100)]
pnv/psi: Add device reset hook
And call it from a QEMU reset handler. This allows each PNV child class to
override the reset hook if needed, eg. POWER8 doesn't but POWER9 does.
The proper way to do that would be to use device_class_set_parent_reset(),
but defining a Pnv8PsiClass and a Pnv9PsiClass types with a parent_reset
pointer adds a fair amount of code. Calling pnv_psi_reset() explicitely is
fine for now.
A subsequent patch will consolidate the call to qemu_register_reset() in
a single place.
Greg Kurz [Mon, 6 Jan 2020 14:56:42 +0000 (15:56 +0100)]
spapr/xive: Deduce the SpaprXive pointer from XiveTCTX::xptr
And use it instead of reaching out to the machine. This allows to get rid
of a call to qdev_get_machine() and to reduce the scope of another one so
that it is only used within the argument list of error_append_hint(). This
is an acceptable tradeoff compared to all it would require to know about
the maximum number of CPUs here without calling qdev_get_machine().
Greg Kurz [Mon, 6 Jan 2020 14:56:38 +0000 (15:56 +0100)]
xive: Use the XIVE fabric link under the XIVE router
Now that the spapr and pnv machines do set the "xive-fabric" link, the
use of the XIVE fabric pointer becomes mandatory. This is checked with
an assert() in a new realize hook. Since the XIVE router is realized at
machine init for the all the machine's life time, no risk to abort an
already running guest (ie. not a hotplug path).
Greg Kurz [Mon, 6 Jan 2020 14:56:37 +0000 (15:56 +0100)]
spapr, pnv, xive: Add a "xive-fabric" link to the XIVE router
In order to get rid of qdev_get_machine(), first add a pointer to the
XIVE fabric under the XIVE router and make it configurable through a
QOM link property.
Configure it in the spapr and pnv machine. In the case of pnv, the XIVE
routers are under the chip, so this is done with a QOM alias property of
the POWER9 pnv chip.
Greg Kurz [Mon, 6 Jan 2020 14:56:36 +0000 (15:56 +0100)]
pnv/xive: Use device_class_set_parent_realize()
The XIVE router base class currently inherits an empty realize hook
from the sysbus device base class, but it will soon implement one
of its own to perform some sanity checks. Do the preliminary plumbing
to have it called.
spapr/xive: remove redundant check in spapr_match_nvt()
spapr_match_nvt() is a XIVE operation and is used by the machine to
look for a matching target when an event notification is being
delivered. An assert checks that spapr_match_nvt() is called only when
the machine has selected the XIVE interrupt mode but it is redundant
with the XIVE_PRESENTER() dynamic cast.
Apply the cast to spapr->active_intc and remove the assert.
Greg Kurz [Mon, 6 Jan 2020 10:29:42 +0000 (11:29 +0100)]
ppc/pnv: Drop "num-chips" machine property
The number of CPU chips of the powernv machine is configurable through a
"num-chips" property. This doesn't fit well with the CPU topology, eg.
some configurations can come up with more CPUs than the maximum of CPUs
set in the toplogy. This causes assertion to be hit with mttcg:
Mttcg mandates the CPU topology to be dimensioned to the actual number
of CPUs, depending on the number of chips the user asked for. That is,
'-machine num-chips=N' should always have a '-smp' companion with a
topology that meats the resulting number of CPUs, typically
'-smp sockets=N'.
It thus seems that "num-chips" doesn't bring anything but forcing the user
to specify the requested number of chips on the command line twice. Simplify
the command line by computing the number of chips based on the CPU topology
exclusively. The powernv machine isn't a production thing ; it is mostly
used by developpers to prepare the bringup of real HW. Because of this and
for simplicity, this deliberately ignores the official deprecation process
and dumps "num-chips" right away : '-smp sockets=N' is now the only way to
control the number of CPU chips.
This is done at machine init because smp_parse() is called after instance
init.
Greg Kurz [Thu, 19 Dec 2019 18:11:47 +0000 (19:11 +0100)]
spapr/xive: Use device_class_set_parent_realize()
The XIVE router base class currently inherits an empty realize hook
from the sysbus device base class, but it will soon implement one
of its own to perform some sanity checks. Do the preliminary plumbing
to have it called.
projects / qemu.git / log