Avoid reusing vector temporaries so that we may re-use them
when propagating stores to loads.
Reviewed-by: Song Gao <gaosong@loongson.cn>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
The movcond opcode is now mandatory for backends to implement.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231026041404.1229328-7-richard.henderson@linaro.org>
The movcond opcode is now mandatory for backends to implement.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231026041404.1229328-4-richard.henderson@linaro.org>
Expand as branch over move if not supported in the ISA.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231026041404.1229328-3-richard.henderson@linaro.org>
Return the temp and a set of flags, to be used as a
primitive for setcond, brcond, movcond.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231026041404.1229328-2-richard.henderson@linaro.org>
These functions are no longer used outside tcg-op.c.
There are several that are completely unused, so remove them.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231029210848.78234-9-richard.henderson@linaro.org>
These are used within tcg-op.c and tcg-op-ldst.c.
There are no uses outside tcg/.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231029210848.78234-8-richard.henderson@linaro.org>
These are used within tcg-op-vec.c and tcg/host/tcg-target.c.inc.
There are no uses outside tcg/.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231029210848.78234-7-richard.henderson@linaro.org>
This one is more complicated, combining 32-bit and 64-bit
expansion with C if instead of preprocessor #if.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231029210848.78234-6-richard.henderson@linaro.org>
In addition to moving out of line, with CONFIG_DEBUG_TCG
mark them all noinline.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231029210848.78234-3-richard.henderson@linaro.org>
Encourage the compiler to tail-call rather than inline
across the dozens of opcode expanders.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231029210848.78234-2-richard.henderson@linaro.org>
Store bytes under a mask is fundamentally a cmpxchg, not a straight store.
Use HAVE_CMPXCHG128 instead of HAVE_ATOMIC128_RW.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230916220151.526140-8-richard.henderson@linaro.org>
We handled the HAVE_ATOMIC128_RW case with atomic16_set at the top of
the function; the only thing left for a host without that support is
to fall through to cpu_loop_exit_atomic.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230916220151.526140-7-richard.henderson@linaro.org>
While loongarch64 does not have a 128-bit cmpxchg, it does
have 128-bit atomic load and store via the vector unit.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230916220151.526140-6-richard.henderson@linaro.org>
Use new registers for the output, so that we never overlap
the input address, which could happen for user-only.
This avoids a "tmp = addr + 0" in that case.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Jiajie Chen <c@jia.je>
Message-Id: <20230916220151.526140-3-richard.henderson@linaro.org>
Constraint with two outputs, both in new registers.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Jiajie Chen <c@jia.je>
Message-Id: <20230916220151.526140-2-richard.henderson@linaro.org>
Move all of it into accel/tcg/monitor.c. This puts everything
about tcg that is only used by the monitor in the same place.
Tested-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
raw_co_zone_append() sets "s->offset" where "BDRVRawState *s". This pointer
is used later at raw_co_prw() to save the block address where the data is
written.
When multiple IOs are on-going at the same time, a later IO's
raw_co_zone_append() call over-writes a former IO's offset address before
raw_co_prw() completes. As a result, the former zone append IO returns the
initial value (= the start address of the writing zone), instead of the
proper address.
Fix the issue by passing the offset pointer to raw_co_prw() instead of
passing it through s->offset. Also, remove "offset" from BDRVRawState as
there is no usage anymore.
Fixes: 4751d09adc ("block: introduce zone append write for zoned devices")
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Message-Id: <20231030073853.2601162-1-naohiro.aota@wdc.com>
Reviewed-by: Sam Li <faithilikerun@gmail.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Hanna Czenczek <hreitz@redhat.com>
When the zoned request fail, it needs to update only the wp of
the target zones for not disrupting the in-flight writes on
these other zones. The wp is updated successfully after the
request completes.
Fixed the callers with right offset and nr_zones.
Signed-off-by: Sam Li <faithilikerun@gmail.com>
Message-Id: <20230825040556.4217-1-faithilikerun@gmail.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
[hreitz: Rebased and fixed comment spelling]
Signed-off-by: Hanna Czenczek <hreitz@redhat.com>
When the discard-no-unref flag is enabled, we keep the reference for
normal discard requests.
But when a discard is executed on a snapshot/qcow2 image with backing,
the discards are saved as zero clusters in the snapshot image.
When committing the snapshot to the backing file, not
discard_in_l2_slice is called but zero_in_l2_slice. Which did not had
any logic to keep the reference when discard-no-unref is enabled.
Therefor we add logic in the zero_in_l2_slice call to keep the reference
on commit.
Fixes: https://gitlab.com/qemu-project/qemu/-/issues/1621
Signed-off-by: Jean-Louis Dupond <jean-louis@dupond.be>
Message-Id: <20231003125236.216473-2-jean-louis@dupond.be>
[hreitz: Made the documentation change more verbose, as discussed
on-list]
Signed-off-by: Hanna Czenczek <hreitz@redhat.com>
In commit be23a049 in the conversion to decodetree we broke the
decoding of the immediate value in the LDRA instruction. This should
be a 10 bit signed value that is scaled by 8, but in the conversion
we incorrectly ended up scaling it only by 2. Fix the scaling
factor.
Cc: qemu-stable@nongnu.org
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1970
Fixes: be23a049 ("target/arm: Convert load (pointer auth) insns to decodetree")
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20231106113445.1163063-1-peter.maydell@linaro.org
On the vexpress-a9 board we try to map both RAM and flash to address 0,
as seen in "info mtree":
address-space: memory
0000000000000000-ffffffffffffffff (prio 0, i/o): system
0000000000000000-0000000003ffffff (prio 0, romd): alias vexpress.flashalias @vexpress.flash0 0000000000000000-0000000003ffffff
0000000000000000-0000000003ffffff (prio 0, ram): alias vexpress.lowmem @vexpress.highmem 0000000000000000-0000000003ffffff
0000000010000000-0000000010000fff (prio 0, i/o): arm-sysctl
0000000010004000-0000000010004fff (prio 0, i/o): pl041
(etc)
The flash "wins" and the RAM mapping is useless (but also harmless).
This happened as a result of commit 6ec1588e in 2014, which changed
"we always map the RAM to the low addresses for vexpress-a9" to "we
always map flash in the low addresses", but forgot to stop mapping
the RAM.
In real hardware, this low part of memory is remappable, both at
runtime by the guest writing to a control register, and configurably
as to what you get out of reset -- you can have the first flash
device, or the second, or the DDR2 RAM, or the external AXI bus
(which for QEMU means "nothing there"). In an ideal world we would
support that remapping both at runtime and via a machine property to
select the out-of-reset behaviour.
Pending anybody caring enough to implement the full remapping
behaviour:
* remove the useless mapped-but-inaccessible lowram MR
* document that QEMU doesn't support remapping of low memory
Fixes: 6ec1588e ("hw/arm/vexpress: Alias NOR flash at 0 for vexpress-a9")
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1761
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20231103185602.875849-1-peter.maydell@linaro.org
For SO_EE_ORIGIN_ZEROCOPY the 32-bit notification range is encoded
as [ee_info, ee_data] inclusively, so ee_info should be less or
equal to ee_data.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Reviewed-by: Maksim Davydov <davydov-max@yandex-team.ru>
Message-id: 20231017125941.810461-7-vsementsov@yandex-team.ru
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Coverity signals that variable as being used uninitialized. And really,
when work with external APIs that's better to zero out the structure,
where we set some fields by hand.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Maksim Davydov <davydov-max@yandex-team.ru>
Message-id: 20231017125941.810461-6-vsementsov@yandex-team.ru
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
NVMeQueuePair::reqs has length NVME_NUM_REQS, which less than
NVME_QUEUE_SIZE by 1.
Fixes: 1086e95da1 ("block/nvme: switch to a NVMeRequest freelist")
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Maksim Davydov <davydov-max@yandex-team.ru>
Message-id: 20231017125941.810461-5-vsementsov@yandex-team.ru
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
set_time() function doesn't set all the fields, so it's better to
initialize tm structure. And Coverity will be happier about it.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Maksim Davydov <davydov-max@yandex-team.ru>
Message-id: 20231017125941.810461-4-vsementsov@yandex-team.ru
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
We support only 3- and 4-level page-tables, which is firstly checked in
vtd_decide_config(), then setup in vtd_init(). Than level fields are
checked by vtd_is_level_supported().
So here we can't have level out from 1..4 inclusive range. Let's assert
it. That also explains Coverity that we are not going to overflow the
array.
CID: 1487158, 1487186
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Maksim Davydov <davydov-max@yandex-team.ru>
Message-id: 20231017125941.810461-2-vsementsov@yandex-team.ru
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Documentation for using the GAS in ACPI tables to report debug UART addresses at
https://learn.microsoft.com/en-us/windows-hardware/drivers/bringup/acpi-debug-port-table
states the following:
- The Register Bit Width field contains the register stride and must be a
power of 2 that is at least as large as the access size. On 32-bit
platforms this value cannot exceed 32. On 64-bit platforms this value
cannot exceed 64.
- The Access Size field is used to determine whether byte, WORD, DWORD, or
QWORD accesses are to be used. QWORD accesses are only valid on 64-bit
architectures.
Documentation for the ARM PL011 at
https://developer.arm.com/documentation/ddi0183/latest/
states that the registers are:
- spaced 4 bytes apart (see Table 3-2), so register stride must be 32.
- 16 bits in size in some cases (see individual registers), so access
size must be at least 2.
Linux doesn't seem to care about this error in the table, but it does
affect at least the NOVA microhypervisor.
In theory we therefore have a choice between reporting the access
size as 2 (16 bit accesses) or 3 (32-bit accesses). In practice,
Linux does not correctly handle the case where the table reports the
access size as 2: as of kernel commit 750b95887e5678, the code in
acpi_parse_spcr() tries to tell the serial driver to use 16 bit
accesses by passing "mmio16" in the option string, but the PL011
driver code in pl011_console_match() only recognizes "mmio" or
"mmio32". The result is that unless the user has enabled 'earlycon'
there is no console output from the guest kernel.
We therefore choose to report the access size as 32 bits; this works
for NOVA and also for Linux. It is also what the UEFI firmware on a
Raspberry Pi 4 reports, so we're in line with existing real-world
practice.
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1938
Signed-off-by: Udo Steinberg <udo@hypervisor.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
[PMM: minor commit message tweaks; use 32 bit accesses]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Allow changes to the virt board SPCR and DBG2 -- we are going to fix
an error in the UART descriptions there.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Since commit 9036e917f8 ("{include/}hw/arm: refactor virt PPI logic")
PMU IRQ registration fails for arm64 guests:
[ 0.563689] hw perfevents: unable to request IRQ14 for ARM PMU counters
[ 0.565160] armv8-pmu: probe of pmu failed with error -22
That commit re-defined VIRTUAL_PMU_IRQ to be a INTID but missed a case
where the PMU IRQ is actually referred by its PPI index. Fix that by using
INTID_TO_PPI() in that case.
Fixes: 9036e917f8 ("{include/}hw/arm: refactor virt PPI logic")
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1960
Signed-off-by: Sebastian Ott <sebott@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 475d918d-ab0e-f717-7206-57a5beb28c7b@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
If we decide to apply this patch (for easier backporting reasons), we
can now revert it.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: Peter Xu <peterx@redhat.com>
Add the necessary plumbing for the hv-balloon driver to the PC machine.
Co-developed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
Used by the hv-balloon driver for (optional) guest memory status reports.
Acked-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
Used by the driver to report its provided memory state information.
Co-developed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
One of advantages of using this protocol over ACPI-based PC DIMM hotplug is
that it allows hot-adding memory in much smaller granularity because the
ACPI DIMM slot limit does not apply.
In order to enable this functionality a new memory backend needs to be
created and provided to the driver via the "memdev" parameter.
This can be achieved by, for example, adding
"-object memory-backend-ram,id=mem1,size=32G" to the QEMU command line and
then instantiating the driver with "memdev=mem1" parameter.
The device will try to use multiple memslots to cover the memory backend in
order to reduce the size of metadata for the not-yet-hot-added part of the
memory backend.
Co-developed-by: David Hildenbrand <david@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
This driver is like virtio-balloon on steroids: it allows both changing the
guest memory allocation via ballooning and (in the next patch) inserting
pieces of extra RAM into it on demand from a provided memory backend.
The actual resizing is done via ballooning interface (for example, via
the "balloon" HMP command).
This includes resizing the guest past its boot size - that is, hot-adding
additional memory in granularity limited only by the guest alignment
requirements, as provided by the next patch.
In contrast with ACPI DIMM hotplug where one can only request to unplug a
whole DIMM stick this driver allows removing memory from guest in single
page (4k) units via ballooning.
After a VM reboot the guest is back to its original (boot) size.
In the future, the guest boot memory size might be changed on reboot
instead, taking into account the effective size that VM had before that
reboot (much like Hyper-V does).
For performance reasons, the guest-released memory is tracked in a few
range trees, as a series of (start, count) ranges.
Each time a new page range is inserted into such tree its neighbors are
checked as candidates for possible merging with it.
Besides performance reasons, the Dynamic Memory protocol itself uses page
ranges as the data structure in its messages, so relevant pages need to be
merged into such ranges anyway.
One has to be careful when tracking the guest-released pages, since the
guest can maliciously report returning pages outside its current address
space, which later clash with the address range of newly added memory.
Similarly, the guest can report freeing the same page twice.
The above design results in much better ballooning performance than when
using virtio-balloon with the same guest: 230 GB / minute with this driver
versus 70 GB / minute with virtio-balloon.
During a ballooning operation most of time is spent waiting for the guest
to come up with newly freed page ranges, processing the received ranges on
the host side (in QEMU and KVM) is nearly instantaneous.
The unballoon operation is also pretty much instantaneous:
thanks to the merging of the ballooned out page ranges 200 GB of memory can
be returned to the guest in about 1 second.
With virtio-balloon this operation takes about 2.5 minutes.
These tests were done against a Windows Server 2019 guest running on a
Xeon E5-2699, after dirtying the whole memory inside guest before each
balloon operation.
Using a range tree instead of a bitmap to track the removed memory also
means that the solution scales well with the guest size: even a 1 TB range
takes just a few bytes of such metadata.
Since the required GTree operations aren't present in every Glib version
a check for them was added to the meson build script, together with new
"--enable-hv-balloon" and "--disable-hv-balloon" configure arguments.
If these GTree operations are missing in the system's Glib version this
driver will be skipped during QEMU build.
An optional "status-report=on" device parameter requests memory status
events from the guest (typically sent every second), which allow the host
to learn both the guest memory available and the guest memory in use
counts.
Following commits will add support for their external emission as
"HV_BALLOON_STATUS_REPORT" QMP events.
The driver is named hv-balloon since the Linux kernel client driver for
the Dynamic Memory Protocol is named as such and to follow the naming
pattern established by the virtio-balloon driver.
The whole protocol runs over Hyper-V VMBus.
The driver was tested against Windows Server 2012 R2, Windows Server 2016
and Windows Server 2019 guests and obeys the guest alignment requirements
reported to the host via DM_CAPABILITIES_REPORT message.
Acked-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
This commit adds Hyper-V Dynamic Memory Protocol definitions, taken
from hv_balloon Linux kernel driver, adapted to the QEMU coding style and
definitions.
Acked-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
Richard Henderson