The lack of SVE memory instrumentation has been an omission in plugin
handling since it was introduced. Fortunately we can utilise the
probe_* functions to force all all memory access to follow the slow
path. We do this by checking the access type and presence of plugin
memory callbacks and if set return the TLB_MMIO flag.
We have to jump through a few hoops in user mode to re-use the flag
but it was the desired effect:
./qemu-system-aarch64 -display none -serial mon:stdio \
-M virt -cpu max -semihosting-config enable=on \
-kernel ./tests/tcg/aarch64-softmmu/memory-sve \
-plugin ./contrib/plugins/libexeclog.so,ifilter=st1w,afilter=0x40001808 -d plugin
gives (disas doesn't currently understand st1w):
0, 0x40001808, 0xe54342a0, ".byte 0xa0, 0x42, 0x43, 0xe5", store, 0x40213010, RAM, store, 0x40213014, RAM, store, 0x40213018, RAM
And for user-mode:
./qemu-aarch64 \
-plugin contrib/plugins/libexeclog.so,afilter=0x4007c0 \
-d plugin \
./tests/tcg/aarch64-linux-user/sha512-sve
gives:
1..10
ok 1 - do_test(&tests[i])
0, 0x4007c0, 0xa4004b80, ".byte 0x80, 0x4b, 0x00, 0xa4", load, 0x5500800370, load, 0x5500800371, load, 0x5500800372, load, 0x5500800373, load, 0x5500800374, load, 0x5500800375, load, 0x5500800376, load, 0x5500800377, load, 0x5500800378, load, 0x5500800379, load, 0x550080037a, load, 0x550080037b, load, 0x550080037c, load, 0x550080037d, load, 0x550080037e, load, 0x550080037f, load, 0x5500800380, load, 0x5500800381, load, 0x5500800382, load, 0x5500800383, load, 0x5500800384, load, 0x5500800385, load, 0x5500800386, lo
ad, 0x5500800387, load, 0x5500800388, load, 0x5500800389, load, 0x550080038a, load, 0x550080038b, load, 0x550080038c, load, 0x550080038d, load, 0x550080038e, load, 0x550080038f, load, 0x5500800390, load, 0x5500800391, load, 0x5500800392, load, 0x5500800393, load, 0x5500800394, load, 0x5500800395, load, 0x5500800396, load, 0x5500800397, load, 0x5500800398, load, 0x5500800399, load, 0x550080039a, load, 0x550080039b, load, 0x550080039c, load, 0x550080039d, load, 0x550080039e, load, 0x550080039f, load, 0x55008003a0, load, 0x55008003a1, load, 0x55008003a2, load, 0x55008003a3, load, 0x55008003a4, load, 0x55008003a5, load, 0x55008003a6, load, 0x55008003a7, load, 0x55008003a8, load, 0x55008003a9, load, 0x55008003aa, load, 0x55008003ab, load, 0x55008003ac, load, 0x55008003ad, load, 0x55008003ae, load, 0x55008003af
(4007c0 is the ld1b in the sha512-sve)
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Cc: Robert Henry <robhenry@microsoft.com>
Cc: Aaron Lindsay <aaron@os.amperecomputing.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230630180423.558337-20-alex.bennee@linaro.org>
We want all accelerators to share the same opaque pointer in
CPUState.
Rename the 'hvf_vcpu_state' structure as 'AccelCPUState'.
Use the generic 'accel' field of CPUState instead of 'hvf'.
Replace g_malloc0() by g_new0() for readability.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20230624174121.11508-17-philmd@linaro.org>
We want all accelerators to share the same opaque pointer in
CPUState. Start with the HAX context, renaming its forward
declarated structure 'hax_vcpu_state' as 'AccelCPUState'.
Document the CPUState field. Directly use the typedef.
Remove the amusing but now unnecessary casts in NVMM / WHPX.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230624174121.11508-8-philmd@linaro.org>
We have run out of bits we can use within the CPUTLBEntry comparators,
as TLB_FLAGS_MASK cannot overlap alignment.
Store slow_flags[] in CPUTLBEntryFull, and merge with the flags from
the comparator. A new TLB_FORCE_SLOW bit is set within the comparator
as an indication that the slow path must be used.
Move TLB_BSWAP to TLB_SLOW_FLAGS_MASK. Since we are out of bits,
we cannot create a new bit without moving an old one.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
A regression has been detected in latency testing of KVM guests.
More specifically, it was observed that the cyclictest
numbers inside of an isolated vcpu (running on isolated pcpu) are:
Where a maximum of 50us is acceptable.
The implementation of KVM_GET_STATS_FD uses run_on_cpu to query
per vcpu statistics, which interrupts the vcpu (and is unnecessary).
To fix this, open the per vcpu stats fd on vcpu initialization,
and read from that fd from QEMU's main thread.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This reverts commit d7ee93e243.
That commit tries to make a field in the CPUState struct not be
present when CONFIG_USER_ONLY is set. Unfortunately, you can't
conditionally omit fields in structs like this based on ifdefs that
are set per-target. If you try it, then code in files compiled
per-target (where CONFIG_USER_ONLY is or can be set) will disagree
about the struct layout with files that are compiled once-only (where
this kind of ifdef is never set).
This manifests specifically in 'make check-tcg' failing, because code
in cpus-common.c that sets up the CPUState::cpu_index field puts it
at a different offset from the code in plugins/core.c in
qemu_plugin_vcpu_init_hook() which reads the cpu_index field. The
latter then hits an assert because from its point of view every
thread has a 0 cpu_index. There might be other weird behaviour too.
Mostly we catch this kind of bug because the CONFIG_whatever is
listed in include/exec/poison.h and so the reference to it in
build-once source files will then cause a compiler error.
Unfortunately CONFIG_USER_ONLY is an exception to that: we have some
places where we use it in "safe" ways in headers that will be seen by
once-only source files (e.g. ifdeffing out function prototypes) and
it would be a lot of refactoring to be able to get to a position
where we could poison it. This leaves us in a "you have to be
careful to walk around the bear trap" situation...
Fixes: d7ee93e243 ("cputlb: Restrict SavedIOTLB to system emulation")
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20230620175712.1331625-1-peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Now we no longer have dynamic state affecting things we can remove the
additional fields in cpu.h and simplify the TB hash calculation.
For the benchmark:
hyperfine -w 2 -m 20 \
"./arm-softmmu/qemu-system-arm -cpu cortex-a15 \
-machine type=virt,highmem=off \
-display none -m 2048 \
-serial mon:stdio \
-netdev user,id=unet,hostfwd=tcp::2222-:22 \
-device virtio-net-pci,netdev=unet \
-device virtio-scsi-pci \
-blockdev driver=raw,node-name=hd,discard=unmap,file.driver=host_device,file.filename=/dev/zen-disk/debian-bullseye-armhf \
-device scsi-hd,drive=hd -smp 4 \
-kernel /home/alex/lsrc/linux.git/builds/arm/arch/arm/boot/zImage \
-append 'console=ttyAMA0 root=/dev/sda2 systemd.unit=benchmark.service' \
-snapshot"
It has a marginal effect on runtime, before:
Time (mean ± σ): 26.279 s ± 2.438 s [User: 41.113 s, System: 1.843 s]
Range (min … max): 24.420 s … 32.565 s 20 runs
after:
Time (mean ± σ): 24.440 s ± 2.885 s [User: 34.474 s, System: 2.028 s]
Range (min … max): 21.663 s … 29.937 s 20 runs
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1358
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20230526165401.574474-10-alex.bennee@linaro.org
Message-Id: <20230524133952.3971948-9-alex.bennee@linaro.org>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
In some spots, it would be helpful to be able to use TARGET_NAME
in common (target independent) code, too. Thus introduce a wrapper
that can be called from common code, too, just like we already
have one for target_words_bigendian().
Message-Id: <20230424160434.331175-3-thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
For being able to create a universal QEMU binary one day, core
files like machine-qmp-cmds.c must not contain any "#ifdef TARGET_..."
parts. Thus let's provide the target specific function via a
function pointer in CPUClass instead, as a first step towards
making this file target independent.
Message-Id: <20230424160434.331175-2-thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
CPU watchpoints can be use by non-TCG accelerators.
KVM uses them:
$ git grep CPUWatchpoint|fgrep kvm
target/arm/kvm64.c:1558: CPUWatchpoint *wp = find_hw_watchpoint(cs, debug_exit->far);
target/i386/kvm/kvm.c:5216:static CPUWatchpoint hw_watchpoint;
target/ppc/kvm.c:443:static CPUWatchpoint hw_watchpoint;
target/s390x/kvm/kvm.c:139:static CPUWatchpoint hw_watchpoint;
See for example commit e4482ab7e3 ("target-arm: kvm - add support
for HW assisted debug"):
This adds basic support for HW assisted debug. The ioctl interface
to KVM allows us to pass an implementation defined number of break
and watch point registers. [...]
This partially reverts commit 2609ec2868.
Fixes: 2609ec2868 ("softmmu: Extract watchpoint API from physmem.c")
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230328173117.15226-4-philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Both cpu_check_watchpoint() and cpu_watchpoint_address_matches()
are specific to TCG system emulation. Declare them in "tcg-cpu-ops.h"
to be sure accessing them from non-TCG code is a compilation error.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230328173117.15226-2-philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Atomic operations are read-modify-write, and we'd like to
be able to test both read and write with one call. This is
easy enough, with BP_MEM_READ | BP_MEM_WRITE.
Add BP_HIT_SHIFT to make it easy to set BP_WATCHPOINT_HIT_*.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Currently dying to one of the core_dump_signal()s deadlocks, because
dump_core_and_abort() calls start_exclusive() two times: first via
stop_all_tasks(), and then via preexit_cleanup() ->
qemu_plugin_user_exit().
There are a number of ways to solve this: resume after dumping core;
check cpu_in_exclusive_context() in qemu_plugin_user_exit(); or make
{start,end}_exclusive() recursive. Pick the last option, since it's
the most straightforward one.
Fixes: da91c19202 ("linux-user: Clean up when exiting due to a signal")
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-Id: <20230214140829.45392-3-iii@linux.ibm.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This API allows the accelerators to prevent vcpus from issuing
new ioctls while execting a critical section marked with the
accel_ioctl_inhibit_begin/end functions.
Note that all functions submitting ioctls must mark where the
ioctl is being called with accel_{cpu_}ioctl_begin/end().
This API requires the caller to always hold the BQL.
API documentation is in sysemu/accel-blocker.h
Internally, it uses a QemuLockCnt together with a per-CPU QemuLockCnt
(to minimize cache line bouncing) to keep avoid that new ioctls
run when the critical section starts, and a QemuEvent to wait
that all running ioctls finish.
Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20221111154758.1372674-2-eesposit@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Replace a config-time define with a compile time condition
define (compatible with clang and gcc) that must be declared prior to
its usage. This avoids having a global configure time define, but also
prevents from bad usage, if the config header wasn't included before.
This can help to make some code independent from qemu too.
gcc supports __BYTE_ORDER__ from about 4.6 and clang from 3.2.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
[ For the s390x parts I'm involved in ]
Acked-by: Halil Pasic <pasic@linux.ibm.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20220323155743.1585078-7-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
One less qemu-specific macro. It also helps to make some headers/units
only depend on glib, and thus moved in standalone projects eventually.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Richard W.M. Jones <rjones@redhat.com>
ArchCPU is our interface with target-specific code. Use it as
a forward-declared opaque pointer (abstract type), having its
structure defined by each target.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20220214183144.27402-15-f4bug@amsat.org>
While CPUState is our interface with generic code, CPUArchState is
our interface with target-specific code. Use CPUArchState as an
abstract type, defined by each target.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20220214183144.27402-13-f4bug@amsat.org>
This reverts commit 1b36e4f5a5.
Despite a comment saying why cpu_common_props cannot be placed in
a file that is compiled once, it was moved anyway. Revert that.
Since then, Property is not defined in hw/core/cpu.h, so it is now
easier to declare a function to install the properties rather than
the Property array itself.
Cc: Eduardo Habkost <ehabkost@redhat.com>
Suggested-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Despite the comment, the members were not kept at the end.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This will allow a breakpoint hack to move out of AVR's translator.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Adjust types for some memory access functions.
Reduce inclusion of tcg headers.
Fix watchpoints vs replay.
Fix tcg/aarch64 roli expansion.
Introduce SysemuCPUOps structure.
# gpg: Signature made Thu 27 May 2021 00:43:54 BST
# gpg: using RSA key 7A481E78868B4DB6A85A05C064DF38E8AF7E215F
# gpg: issuer "richard.henderson@linaro.org"
# gpg: Good signature from "Richard Henderson <richard.henderson@linaro.org>" [full]
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A 05C0 64DF 38E8 AF7E 215F
* remotes/rth-gitlab/tags/pull-tcg-20210526: (31 commits)
hw/core: Constify TCGCPUOps
target/mips: Fold jazz behaviour into mips_cpu_do_transaction_failed
cpu: Move CPUClass::get_paging_enabled to SysemuCPUOps
cpu: Move CPUClass::get_memory_mapping to SysemuCPUOps
cpu: Move CPUClass::get_phys_page_debug to SysemuCPUOps
cpu: Move CPUClass::asidx_from_attrs to SysemuCPUOps
cpu: Move CPUClass::write_elf* to SysemuCPUOps
cpu: Move CPUClass::get_crash_info to SysemuCPUOps
cpu: Move CPUClass::virtio_is_big_endian to SysemuCPUOps
cpu: Move CPUClass::vmsd to SysemuCPUOps
cpu: Introduce SysemuCPUOps structure
cpu: Move AVR target vmsd field from CPUClass to DeviceClass
cpu: Rename CPUClass vmsd -> legacy_vmsd
cpu: Assert DeviceClass::vmsd is NULL on user emulation
cpu: Directly use get_memory_mapping() fallback handlers in place
cpu: Directly use get_paging_enabled() fallback handlers in place
cpu: Directly use cpu_write_elf*() fallback handlers in place
cpu: Introduce cpu_virtio_is_big_endian()
cpu: Un-inline cpu_get_phys_page_debug and cpu_asidx_from_attrs
cpu: Split as cpu-common / cpu-sysemu
...
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
