Peter Maydell [Tue, 13 Oct 2020 10:35:32 +0000 (11:35 +0100)]
target/arm: AArch32 VCVT fixed-point to float is always round-to-nearest
For AArch32, unlike the VCVT of integer to float, which honours the
rounding mode specified by the FPSCR, VCVT of fixed-point to float is
always round-to-nearest. (AArch64 fixed-point-to-float conversions
always honour the FPCR rounding mode.)
Implement this by providing _round_to_nearest versions of the
relevant helpers which set the rounding mode temporarily when making
the call to the underlying softfloat function.
We only need to change the VFP VCVT instructions, because the
standard- FPSCR value used by the Neon VCVT is always set to
round-to-nearest, so we don't need to do the extra work of saving
and restoring the rounding mode.
Peter Maydell [Fri, 9 Oct 2020 14:47:12 +0000 (15:47 +0100)]
target/arm: Fix SMLAD incorrect setting of Q bit
The SMLAD instruction is supposed to:
* signed multiply Rn[15:0] * Rm[15:0]
* signed multiply Rn[31:16] * Rm[31:16]
* perform a signed addition of the products and Ra
* set Rd to the low 32 bits of the theoretical
infinite-precision result
* set the Q flag if the sign-extension of Rd
would differ from the infinite-precision result
(ie on overflow)
Our current implementation doesn't quite do this, though: it performs
an addition of the products setting Q on overflow, and then it adds
Ra, again possibly setting Q. This sometimes incorrectly sets Q when
the architecturally mandated only-check-for-overflow-once algorithm
does not. For instance:
r1 = 0x80008000; r2 = 0x80008000; r3 = 0xffffffff
smlad r0, r1, r2, r3
This is (-32768 * -32768) + (-32768 * -32768) - 1
The products are both 0x4000_0000, so when added together as 32-bit
signed numbers they overflow (and QEMU sets Q), but because the
addition of Ra == -1 brings the total back down to 0x7fff_ffff
there is no overflow for the complete operation and setting Q is
incorrect.
Fix this edge case by resorting to 64-bit arithmetic for the
case where we need to add three values together.
John Snow [Tue, 6 Oct 2020 23:58:09 +0000 (19:58 -0400)]
python/qemu/console_socket.py: Correct type of recv()
The type and parameter names of recv() should match socket.socket().
OK, easy enough, but in the cases we don't pass straight through to the
real socket implementation, we probably can't accept such flags. OK, for
now, assert that we don't receive flags in such cases.
John Snow [Tue, 6 Oct 2020 23:58:08 +0000 (19:58 -0400)]
python/qemu: Add mypy type annotations
These should all be purely annotations with no changes in behavior at
all. You need to be in the python folder, but you should be able to
confirm that these annotations are correct (or at least self-consistent)
by running `mypy --strict qemu`.
John Snow [Tue, 6 Oct 2020 23:58:07 +0000 (19:58 -0400)]
iotests.py: Adjust HMP kwargs typing
mypy wants to ensure there's consistency between the kwargs arguments
types and any unspecified keyword arguments. In this case, conv_keys is
a bool, but the remaining keys are Any type. Mypy (correctly) infers the
**kwargs type to be **Dict[str, str], which is not compatible with
conv_keys: bool.
Because QMP typing is a little fraught right now anyway, re-type kwargs
to Dict[str, Any] which has the benefit of silencing this check right
now.
A future re-design might type these more aggressively, but this will
give us a baseline to work from with minimal disruption.
John Snow [Tue, 6 Oct 2020 23:58:06 +0000 (19:58 -0400)]
python/qemu: make 'args' style arguments immutable
These arguments don't need to be mutable and aren't really used as
such. Clarify their types as immutable and adjust code to match where
necessary.
In general, It's probably best not to accept a user-defined mutable
object and store it as internal object state unless there's a strong
justification for doing so. Instead, try to use generic types as input
with empty tuples as the default, and coerce to list where necessary.
John Snow [Tue, 6 Oct 2020 23:58:04 +0000 (19:58 -0400)]
python/machine.py: Add _qmp access shim
Like many other Optional[] types, it's not always a given that this
object will be set. Wrap it in a type-shim that raises a meaningful
error and will always return a concrete type.
John Snow [Tue, 6 Oct 2020 23:58:01 +0000 (19:58 -0400)]
python/machine.py: Don't modify state in _base_args()
Don't append to the _remove_files list during _base_args; instead do so
during _launch. Rework _base_args as a @property to help facilitate
this impression.
This has the additional benefit of making the type of _console_address
easier to analyze statically.
John Snow [Tue, 6 Oct 2020 23:58:00 +0000 (19:58 -0400)]
python/machine.py: reorder __init__
Put the init arg handling all at the top, and mostly in order (deviating
when one is dependent on another), and put what is effectively runtime
state declaration at the bottom.
John Snow [Tue, 6 Oct 2020 23:57:59 +0000 (19:57 -0400)]
python/machine.py: Fix monitor address typing
Prior to this, it's difficult for mypy to intuit what the concrete type
of the monitor address is; it has difficulty inferring the type across
two variables.
Create _monitor_address as a property that always returns a valid
address to simplify static type analysis.
To preserve our ability to clean up, use a simple boolean to indicate
whether or not we should try to clean up the sock file after execution.
John Snow [Fri, 10 Jul 2020 19:30:14 +0000 (15:30 -0400)]
MAINTAINERS: Add Python library stanza
I'm proposing that I split the actual Python library off from the other
miscellaneous python scripts we have and declare it maintained. Add
myself as a maintainer of this folder, along with Cleber.
I will be actively working to add CI style guide checks, strict typing,
and an actual package infrastructure to this folder specifically which
differentiates it from loose, miscellaneous scripts which are generally
maintained by other individuals with subject matter expertise.
Peter Maydell [Tue, 20 Oct 2020 10:20:36 +0000 (11:20 +0100)]
Merge remote-tracking branch 'remotes/aperard/tags/pull-xen-20201020' into staging
Xen queue
* cleanup patches.
* improve xen backend setup performance when other xen guests are
running/booting.
* improve xen guest migration when running in a stubdomain.
# gpg: Signature made Tue 20 Oct 2020 10:55:11 BST
# gpg: using RSA key F80C006308E22CFD8A92E7980CF5572FD7FB55AF
# gpg: issuer "[email protected]"
# gpg: Good signature from "Anthony PERARD <[email protected]>" [marginal]
# gpg: aka "Anthony PERARD <[email protected]>" [marginal]
# 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: 5379 2F71 024C 600F 778A 7161 D8D5 7199 DF83 42C8
# Subkey fingerprint: F80C 0063 08E2 2CFD 8A92 E798 0CF5 572F D7FB 55AF
* remotes/aperard/tags/pull-xen-20201020:
hw/xen: Set suppress-vmdesc for Xen machines
xen-bus: reduce scope of backend watch
xen: Rename XENBACKEND_DEVICE to XENBACKEND
xen: xenguest is not used so is not needed
Jason Andryuk [Tue, 13 Oct 2020 19:05:06 +0000 (15:05 -0400)]
hw/xen: Set suppress-vmdesc for Xen machines
xen-save-devices-state doesn't currently generate a vmdesc, so restore
always triggers "Expected vmdescription section, but got 0". This is
not a problem when restore comes from a file. However, when QEMU runs
in a linux stubdom and comes over a console, EOF is not received. This
causes a delay restoring - though it does restore.
Setting suppress-vmdesc skips looking for the vmdesc during restore and
avoids the wait.
The other approach would be generate a vmdesc in qemu_save_device_state.
Since COLO shared that function, and the vmdesc is just discarded on
restore, we choose to skip it.
Paul Durrant [Thu, 1 Oct 2020 08:15:00 +0000 (09:15 +0100)]
xen-bus: reduce scope of backend watch
Currently a single watch on /local/domain/X/backend is registered by each
QEMU process running in service domain X (where X is usually 0). The purpose
of this watch is to ensure that QEMU is notified when the Xen toolstack
creates a new device backend area.
Such a backend area is specific to a single frontend area created for a
specific guest domain and, since each QEMU process is also created to service
a specfic guest domain, it is unnecessary and inefficient to notify all QEMU
processes.
Only the QEMU process associated with the same guest domain need
receive the notification. This patch re-factors the watch registration code
such that notifications are targetted appropriately.
Peter Maydell [Mon, 19 Oct 2020 13:39:26 +0000 (14:39 +0100)]
Merge remote-tracking branch 'remotes/cschoenebeck/tags/pull-9p-20201019' into staging
9pfs: add tests using local fs driver
The currently existing 9pfs test cases are all solely using the 9pfs 'synth'
fileystem driver, which is a very simple and purely simulated (in RAM only)
filesystem. There are issues though where the 'synth' fs driver is not
sufficient. For example the following two bugs need test cases running the
9pfs 'local' fs driver:
This patch set for that reason introduces 9pfs test cases using the 9pfs
'local' filesystem driver along to the already existing tests on 'synth'.
# gpg: Signature made Mon 19 Oct 2020 13:39:08 BST
# gpg: using RSA key 96D8D110CF7AF8084F88590134C2B58765A47395
# gpg: issuer "[email protected]"
# gpg: Good signature from "Christian Schoenebeck <[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: ECAB 1A45 4014 1413 BA38 4926 30DB 47C3 A012 D5F4
# Subkey fingerprint: 96D8 D110 CF7A F808 4F88 5901 34C2 B587 65A4 7395
* remotes/cschoenebeck/tags/pull-9p-20201019:
tests/9pfs: add local Tmkdir test
tests/9pfs: add virtio_9p_test_path()
tests/9pfs: wipe local 9pfs test directory
tests/9pfs: introduce local tests
tests/9pfs: change qtest name prefix to synth
9pfs: suppress performance warnings on qtest runs
This test case uses the 9pfs 'local' driver to create a directory
and then checks if the expected directory was actually created
(as real directory) on host side.
This patch introduces a custom split() implementation, because
the test code requires non empty array elements as result. For
that reason g_strsplit() would not be a good alternative, as
it would require additional filter code for reshuffling the
array, and the resulting code would be even more complex than
this split() function.
This new public function virtio_9p_test_path() allows 9pfs
'local' tests to translate a path from guest scope to host
scope. For instance by passing an empty string it would
return the root path on host of the exported 9pfs tree.
Before running the first 9pfs test case, make sure the test directory
for running the 9pfs 'local' tests on is entirely empty. For that
reason simply delete the test directory (if any) before (re)creating
it on test suite startup.
Note: The preferable precise behaviour would be the test directory
only being wiped once *before* a test suite run. Right now the test
directory is also wiped at the *end* of a test suite run because
libqos is calling the virtio_9p_register_nodes() callback for some
reason also when a test suite completed. This is suboptimal as
developers cannot immediately see what files and directories the
9pfs local tests created precisely after the test suite completed.
But fortunately the test directory is not wiped if some test failed.
So it is probably not worth it drilling another hole into libqos
for this issue.
This patch introduces 9pfs test cases using the 9pfs 'local'
filesystem driver which reads/writes/creates/deletes real files
and directories.
In this initial version, there is only one local test which actually
only checks if the 9pfs 'local' device was created successfully.
Before the 9pfs 'local' tests are run, a test directory 'qtest-9p-local'
is created (with world rwx permissions) under the current working
directory. At this point that test directory is not auto deleted yet.
All existing 9pfs test cases are using the 'synth' fs driver so far, which
means they are not accessing real files, but a purely simulated (in RAM
only) file system.
Let's make this clear by changing the prefix of the individual qtest case
names from 'fs/' to 'synth/'. That way they'll be easily distinguishable
from upcoming new 9pfs test cases supposed to be using a different fs
driver.
Peter Maydell [Mon, 19 Oct 2020 10:46:03 +0000 (11:46 +0100)]
Merge remote-tracking branch 'remotes/mcayland/tags/qemu-macppc-20201019' into staging
qemu-macppc updates
# gpg: Signature made Mon 19 Oct 2020 08:13:16 BST
# gpg: using RSA key CC621AB98E82200D915CC9C45BC2C56FAE0F321F
# gpg: issuer "[email protected]"
# gpg: Good signature from "Mark Cave-Ayland <[email protected]>" [full]
# Primary key fingerprint: CC62 1AB9 8E82 200D 915C C9C4 5BC2 C56F AE0F 321F
* remotes/mcayland/tags/qemu-macppc-20201019:
mac_oldworld: Change PCI address of macio to match real hardware
mac_oldworld: Drop some variables
mac_oldworld: Drop a variable, use get_system_memory() directly
mac_newworld: Allow loading binary ROM image
mac_oldworld: Allow loading binary ROM image
m48t59: remove legacy m48t59_init() function
ppc405_boards: use qdev properties instead of legacy m48t59_init() function
sun4u: use qdev properties instead of legacy m48t59_init() function
sun4m: use qdev properties instead of legacy m48t59_init() function
m48t59-isa: remove legacy m48t59_init_isa() function
uninorth: use qdev gpios for PCI IRQs
grackle: use qdev gpios for PCI IRQs
macio: don't reference serial_hd() directly within the device
Peter Maydell [Mon, 19 Oct 2020 09:52:56 +0000 (10:52 +0100)]
Merge remote-tracking branch 'remotes/philmd-gitlab/tags/mips-next-20201017' into staging
MIPS patches queue
. Fix some comment spelling errors
. Demacro some TCG helpers
. Add loongson-ext lswc2/lsdc2 group of instructions
. Log unimplemented cache opcode
. Increase number of TLB entries on the 34Kf core
. Allow the CPU to use dynamic frequencies
. Calculate the CP0 timer period using the CPU frequency
. Set CPU frequency for each machine
. Fix Malta FPGA I/O region size
. Allow running qtests when ROM is missing
. Add record/replay acceptance tests
. Update MIPS CPU documentation
. MAINTAINERS updates
CI jobs results:
https://gitlab.com/philmd/qemu/-/pipelines/203931842
https://travis-ci.org/github/philmd/qemu/builds/736491461
https://cirrus-ci.com/build/6272264062631936
https://app.shippable.com/github/philmd/qemu/runs/886/summary/console
# gpg: Signature made Sat 17 Oct 2020 14:59:53 BST
# gpg: using RSA key FAABE75E12917221DCFD6BB2E3E32C2CDEADC0DE
# gpg: Good signature from "Philippe Mathieu-Daudé (F4BUG) <[email protected]>" [full]
# Primary key fingerprint: FAAB E75E 1291 7221 DCFD 6BB2 E3E3 2C2C DEAD C0DE
* remotes/philmd-gitlab/tags/mips-next-20201017: (44 commits)
target/mips: Increase number of TLB entries on the 34Kf core (16 -> 64)
MAINTAINERS: Remove duplicated Malta test entries
MAINTAINERS: Downgrade MIPS Boston to 'Odd Fixes', fix Paul Burton mail
MAINTAINERS: Put myself forward for MIPS target
MAINTAINERS: Remove myself
docs/system: Update MIPS CPU documentation
tests/acceptance: Add MIPS record/replay tests
hw/mips: Remove exit(1) in case of missing ROM
hw/mips: Rename TYPE_MIPS_BOSTON to TYPE_BOSTON
hw/mips: Simplify code using ROUND_UP(INITRD_PAGE_SIZE)
hw/mips: Simplify loading 64-bit ELF kernels
hw/mips/malta: Use clearer qdev style
hw/mips/malta: Move gt64120 related code together
hw/mips/malta: Fix FPGA I/O region size
target/mips/cpu: Display warning when CPU is used without input clock
hw/mips/cps: Do not allow use without input clock
hw/mips/malta: Set CPU frequency to 320 MHz
hw/mips/boston: Set CPU frequency to 1 GHz
hw/mips/cps: Expose input clock and connect it to CPU cores
hw/mips/jazz: Correct CPU frequencies
...
hw/usb/hcd-dwc2: fix divide-by-zero in dwc2_handle_packet()
Check the value of mps to avoid potential divide-by-zero later in the function.
Since HCCHAR_MPS is guest controllable, this prevents a malicious/buggy guest
from crashing the QEMU process on the host.
Anthony PERARD [Wed, 14 Oct 2020 10:41:06 +0000 (11:41 +0100)]
usb/hcd-ehci: Fix error handling on missing device for iTD
The EHCI Host Controller emulation attempt to locate the device
associated with a periodic isochronous transfer description (iTD) and
when this fail the host controller is reset.
But according the EHCI spec 1.0 section 5.15.2.4 Host System
Error, the host controller is supposed to reset itself only when it
failed to communicate with the Host (Operating System), like when
there's an error on the PCI bus. If a transaction fails, there's
nothing in the spec that say to reset the host controller.
This patch rework the error path so that the host controller can keep
working when the OS setup a bogus transaction, it also revert to the
behavior of the EHCI emulation to before commits: e94682f1fe ("ehci: check device is not NULL before calling usb_ep_get()") 7011baece2 ("usb: remove unnecessary NULL device check from usb_ep_get()")
The issue has been found while trying to passthrough a USB device to a
Windows Server 2012 Xen guest via "usb-ehci", which prevent the USB
device from working in Windows. ("usb-ehci" alone works, windows only
setup this weird periodic iTD to device 127 endpoint 15 when the USB
device is passthrough.)
Paul Zimmerman [Sun, 20 Sep 2020 02:14:49 +0000 (19:14 -0700)]
usb: hcd-dwc2: change assert()s to qemu_log_mask(LOG_GUEST_ERROR...)
Change several assert()s to qemu_log_mask(LOG_GUEST_ERROR...),
to prevent the guest from causing Qemu to assert. Also fix up
several existing qemu_log_mask()s to include the function name in
the message.
BALATON Zoltan [Thu, 15 Oct 2020 23:47:17 +0000 (01:47 +0200)]
mac_oldworld: Change PCI address of macio to match real hardware
The board firmware expect these to be at fixed addresses and programs
them without probing, this patch puts the macio device at the expected
PCI address.
BALATON Zoltan [Thu, 15 Oct 2020 23:47:17 +0000 (01:47 +0200)]
mac_oldworld: Drop some variables
Values not used frequently enough may not worth putting in a local
variable, especially with names almost as long as the original value
because that does not improve readability, to the contrary it makes it
harder to see what value is used. Drop a few such variables.
BALATON Zoltan [Thu, 15 Oct 2020 23:47:17 +0000 (01:47 +0200)]
mac_oldworld: Drop a variable, use get_system_memory() directly
Half of the occurances already use get_system_memory() directly
instead of sysmem variable, convert the two other uses to
get_system_memory() too which seems to be more common and drop the
variable.
BALATON Zoltan [Thu, 15 Oct 2020 23:47:17 +0000 (01:47 +0200)]
mac_newworld: Allow loading binary ROM image
Fall back to load binary ROM image if loading ELF fails. This also
moves PROM_BASE and PROM_SIZE defines to board as these are matching
the ROM size and address on this board and removes the now unused
PROM_ADDR and BIOS_SIZE defines from common mac.h.
BALATON Zoltan [Sat, 17 Oct 2020 15:47:29 +0000 (17:47 +0200)]
mac_oldworld: Allow loading binary ROM image
The beige G3 Power Macintosh has a 4MB firmware ROM. Fix the size of
the rom region and fall back to loading a binary image with -bios if
loading ELF image failed. This allows testing emulation with a ROM
image from real hardware as well as using an ELF OpenBIOS image.
Mark Cave-Ayland [Tue, 13 Oct 2020 11:49:22 +0000 (12:49 +0100)]
uninorth: use qdev gpios for PCI IRQs
Currently an object link property is used to pass a reference to the OpenPIC
into the PCI host bridge so that pci_unin_init_irqs() can connect the PCI
IRQs to the PIC itself.
This can be simplified by defining the PCI IRQs as qdev gpios and then wiring
up the PCI IRQs to the PIC in the New World machine init function.
Mark Cave-Ayland [Tue, 13 Oct 2020 11:49:21 +0000 (12:49 +0100)]
grackle: use qdev gpios for PCI IRQs
Currently an object link property is used to pass a reference to the Heathrow
PIC into the PCI host bridge so that grackle_init_irqs() can connect the PCI
IRQs to the PIC itself.
This can be simplified by defining the PCI IRQs as qdev gpios and then wiring
up the PCI IRQs to the PIC in the Old World machine init function.
* remotes/bonzini-gitlab/tags/for-upstream: (22 commits)
ci: include configure and meson logs in all jobs if configure fails
hax: unbreak accelerator cpu code after cpus.c split
fuzz: Disable QEMU's SIG{INT,HUP,TERM} handlers
cirrus: Enable doc build on msys2/mingw
meson: Move the detection logic for sphinx to meson
meson: move SPHINX_ARGS references within "if build_docs"
docs: Fix Sphinx configuration for msys2/mingw
meson: Only install icons and qemu.desktop if have_system
configure: fix handling of --docdir parameter
meson: cleanup curses/iconv test
meson.build: don't condition iconv detection on library detection
build: add --enable/--disable-libudev
build: replace ninjatool with ninja
build: cleanups to Makefile
add ninja to dockerfiles, CI configurations and test VMs
dockerfiles: enable Centos 8 PowerTools
configure: move QEMU_INCLUDES to meson
tests: add missing generated sources to testqapi
make: run shell with pipefail
tests/Makefile.include: unbreak non-tcg builds
...
Prior to this patch, the only way I found to terminate the fuzzer was
either to:
1. Explicitly specify the number of fuzzer runs with the -runs= flag
2. SIGKILL the process with "pkill -9 qemu-fuzz-*" or similar
In addition to being annoying to deal with, SIGKILLing the process skips
over any exit handlers(e.g. registered with atexit()). This is bad,
since some fuzzers might create temporary files that should ideally be
removed on exit using an exit handler. The only way to achieve a clean
exit now is to specify -runs=N , but the desired "N" is tricky to
identify prior to fuzzing.
Why doesn't the process exit with standard SIGINT,SIGHUP,SIGTERM
signals? QEMU installs its own handlers for these signals in
os-posix.c:os_setup_signal_handling, which notify the main loop that an
exit was requested. The fuzzer, however, does not run qemu_main_loop,
which performs the main_loop_should_exit() check. This means that the
fuzzer effectively ignores these signals. As we don't really care about
cleanly stopping the disposable fuzzer "VM", this patch uninstalls
QEMU's signal handlers. Thus, we can stop the fuzzer with
SIG{INT,HUP,TERM} and the fuzzing code can optionally use atexit() to
clean up temporary files/resources.
Bruce Rogers [Thu, 15 Oct 2020 19:07:42 +0000 (13:07 -0600)]
configure: fix handling of --docdir parameter
Commit ca8c0909f01 changed qemu_docdir to be docdir, then later uses the
qemu_docdir name in the final assignment. Unfortunately, one instance of
qemu_docdir was missed: the one which comes from the --docdir parameter.
This patch restores the proper handling of the --docdir parameter.
Fixes: ca8c0909f01 ("configure: build docdir like other suffixed
directories")
Bruce Rogers [Wed, 14 Oct 2020 22:19:39 +0000 (16:19 -0600)]
meson.build: don't condition iconv detection on library detection
It isn't necessarily the case that use of iconv requires an additional
library. For that reason we shouldn't conditionalize iconv detection on
libiconv.found.
Paolo Bonzini [Thu, 15 Oct 2020 10:09:27 +0000 (06:09 -0400)]
build: add --enable/--disable-libudev
Initially, libudev detection was bundled with --enable-mpath because
qemu-pr-helper was the only user of libudev. Recently however the USB
U2F emulation has also started using libudev, so add a separate
option. This also allows 1) disabling libudev if desired for static
builds and 2) for non-static builds, requiring libudev even if
multipath support is undesirable.
The multipath test is adjusted, because it is now possible to enter it
with configurations that should fail, such as --static --enable-mpath
--disable-libudev.
Paolo Bonzini [Thu, 13 Aug 2020 13:28:11 +0000 (09:28 -0400)]
build: replace ninjatool with ninja
Now that the build is done entirely by Meson, there is no need
to keep the Makefile conversion. Instead, we can ask Ninja about
the targets it exposes and forward them.
The main advantages are, from smallest to largest:
- reducing the possible namespace pollution within the Makefile
- removal of a relatively large Python program
- faster build because parsing Makefile.ninja is slower than
parsing build.ninja; and faster build after Meson runs because
we do not have to generate Makefile.ninja.
- tracking of command lines, which provides more accurate rebuilds
In addition the change removes the requirement for GNU make 3.82, which
was annoying on Mac, and avoids bugs on Windows due to ninjatool not
knowing how to convert Windows escapes to POSIX escapes.
Paolo Bonzini [Wed, 14 Oct 2020 12:45:42 +0000 (08:45 -0400)]
configure: move QEMU_INCLUDES to meson
Confusingly, QEMU_INCLUDES is not used by configure tests. Moving
it to meson.build ensures that Windows paths are specified instead of
the msys paths like /c/Users/...
Greg Kurz [Thu, 15 Oct 2020 14:49:06 +0000 (16:49 +0200)]
Makefile: Ensure cscope.out/tags/TAGS are generated in the source tree
Tools usually expect the index files to be in the source tree, eg. emacs.
This is already the case when doing out-of-tree builds, but with in-tree
builds they end up in the build directory.
Force cscope, ctags and etags to put them in the source tree.
target/mips: Increase number of TLB entries on the 34Kf core (16 -> 64)
Per "MIPS32 34K Processor Core Family Software User's Manual,
Revision 01.13" page 8 in "Joint TLB (JTLB)" section:
"The JTLB is a fully associative TLB cache containing 16, 32,
or 64-dual-entries mapping up to 128 virtual pages to their
corresponding physical addresses."
There is no particular reason to restrict the 34Kf core model to
16 TLB entries, so raise its config to 64.
This is helpful for other projects, in particular the Yocto Project:
Yocto Project uses qemu-system-mips 34Kf cpu model, to run 32bit
MIPS CI loop. It was observed that in this case CI test execution
time was almost twice longer than 64bit MIPS variant that runs
under MIPS64R2-generic model. It was investigated and concluded
that the difference in number of TLBs 16 in 34Kf case vs 64 in
MIPS64R2-generic is responsible for most of CI real time execution
difference. Because with 16 TLBs linux user-land trashes TLB more
and it needs to execute more instructions in TLB refill handler
calls, as result it runs much longer.
MAINTAINERS: Downgrade MIPS Boston to 'Odd Fixes', fix Paul Burton mail
Paul's Wavecomp email has been bouncing for months. He told us
he "no longer has access to modern MIPS CPUs or Boston hardware,
and wouldn't currently have time to spend on them if he did." [1]
but "perhaps that might change in the future." [2].
Be fair and downgrade the status of the Boston board to "Odd Fixes"
(has a maintainer but they don't have time to do much other).
Similarly to commit 2b107c2c1c (".mailmap: Update Paul Burton email
address"), update his email address here too.
I have been working on project other than QEMU for some time, and would
like to devote myself to that project. It is impossible for me to find
enough time to perform maintainer's duties with needed meticulousness
and patience.
I wish prosperous future to QEMU and all colleagues in QEMU community.
Pavel Dovgalyuk [Thu, 15 Oct 2020 11:25:02 +0000 (14:25 +0300)]
tests/acceptance: Add MIPS record/replay tests
This patch adds MIPS-targeted acceptance tests for
record/replay functions.
Signed-off-by: Pavel Dovgalyuk <[email protected]> Reviewed-by: Philippe Mathieu-Daudé <[email protected]> Tested-by: Philippe Mathieu-Daudé <[email protected]>
Message-Id: <160276110297.2705.10918105269658307206.stgit@pasha-ThinkPad-X280>
[PMD: Moved 'override timeout' comment from instance to class,
moved nanomips tests to ReplayKernelSlow class,
tagged ReplayKernelSlow class with AVOCADO_TIMEOUT_EXPECTED] Signed-off-by: Philippe Mathieu-Daudé <[email protected]>
Pavel Dovgalyuk [Tue, 21 Jul 2020 06:15:05 +0000 (09:15 +0300)]
hw/mips: Remove exit(1) in case of missing ROM
This patch updates MIPS-based machines to allow starting them without ROM.
In this case CPU starts to execute instructions from the empty memory,
but QEMU allows introspecting the machine configuration.
hw/mips: Simplify code using ROUND_UP(INITRD_PAGE_SIZE)
Instead of using a INITRD_PAGE_MASK definition, use the
simpler INITRD_PAGE_SIZE one which allows us to simplify
the code by using directly the self-explicit ROUND_UP()
macro.
Since 82790064116 ("Cast ELF datatypes properly to host 64bit types")
we don't need to sign-extend the entry_point address. Remove this
unnecessary code.
target/mips/cpu: Display warning when CPU is used without input clock
All our QOM users provides an input clock. In order to avoid
avoid future machines added without clock, display a warning.
User-mode emulation use the CP0 timer with the RDHWR instruction
(see commit cdfcad788394) so keep using the fixed 200 MHz clock
without diplaying any warning. Only display it in system-mode
emulation.
The CPU frequency is normally provided by the firmware in the
"cpuclock" environment variable. The 2E board can handles up
to 660MHz, but be conservative and take the same value used
by the Linux kernel: 533 MHz.
projects / qemu.git / log