Backport subject : PMU improvements and Perf sampling in KVM guest

[zhuzhenxxx-collab]

patch合入情况：

### 依赖patch：
https://lore.kernel.org/all/cover.1705916069.git.haibo1.xu@intel.com/
[PATCH v5 01/12] selftests/kvm: Fix issues with $(SPLIT_TESTS)  本次合入
[PATCH v5 02/12] KVM: arm64: selftests: Data type cleanup for arch_timer test 未合入
[PATCH v5 03/12] KVM: arm64: selftests: Enable tuning of error margin in " 未合入
[PATCH v5 04/12] KVM: arm64: selftests: Split arch_timer test code 本次合入
[PATCH v5 05/12] KVM: selftests: Add CONFIG_64BIT definition for the build 未合入
[PATCH v5 06/12] tools: riscv: Add header file csr.h 未合入
[PATCH v5 07/12] tools: riscv: Add header file vdso/processor.h 未合入
[PATCH v5 08/12] KVM: riscv: selftests: Switch to use macro from csr.h 未合入
[PATCH v5 09/12] KVM: riscv: selftests: Add exception handling support 本次合入
[PATCH v5 10/12] KVM: riscv: selftests: Add guest helper to get vcpu id 本次合入
[PATCH v5 11/12] KVM: riscv: selftests: Change vcpu_has_ext to a common function 本次合入
[PATCH v5 12/12] KVM: riscv: selftests: Add sstc timer test 本次合入

### 主要功能patch：
https://lore.kernel.org/all/20240420151741.962500-2-atishp@rivosinc.com/ 
[PATCH v8 01/24] RISC-V: Fix the typo in Scountovf CSR name 本次合入
[PATCH v8 02/24] RISC-V: Add FIRMWARE_READ_HI definition 本次合入
[PATCH v8 03/24] drivers/perf: riscv: Read upper bits of a firmware counter 本次合入
[PATCH v8 04/24] drivers/perf: riscv: Use BIT macro for shifting operations 本次合入
[PATCH v8 05/24] RISC-V: Add SBI PMU snapshot definitions 本次合入
[PATCH v8 06/24] RISC-V: KVM: Rename the SBI_STA_SHMEM_DISABLE to a generic name 本次合入
[PATCH v8 07/24] RISC-V: Use the minor version mask while computing sbi version 本次合入
[PATCH v8 08/24] drivers/perf: riscv: Fix counter mask iteration for RV32 本次合入
[PATCH v8 09/24] drivers/perf: riscv: Implement SBI PMU snapshot function 本次合入
[PATCH v8 10/24] RISC-V: KVM: Fix the initial sample period value    (rvck已合入）
[PATCH v8 11/24] RISC-V: KVM: No need to update the counter value during reset 本次合入
[PATCH v8 12/24] RISC-V: KVM: No need to exit to the user space if perf event failed 本次合入
[PATCH v8 13/24] RISC-V: KVM: Implement SBI PMU Snapshot feature 本次合入
[PATCH v8 14/24] RISC-V: KVM: Add perf sampling support for guests 本次合入 
[PATCH v8 15/24] RISC-V: KVM: Support 64 bit firmware counters on RV32 本次合入
[PATCH v8 16/24] RISC-V: KVM: Improve firmware counter read function 本次合入
[PATCH v8 17/24] KVM: riscv: selftests: Move sbi definitions to its own header file 本次合入
[PATCH v8 18/24] KVM: riscv: selftests: Add helper functions for extension checks 本次合入
[PATCH v8 19/24] KVM: riscv: selftests: Add Sscofpmf to get-reg-list test 本次合入
[PATCH v8 20/24] KVM: riscv: selftests: Add SBI PMU extension definitions 本次合入
[PATCH v8 21/24] KVM: riscv: selftests: Add SBI PMU selftest 本次合入
[PATCH v8 22/24] KVM: riscv: selftests: Add a test for PMU snapshot functionality 本次合入
[PATCH v8 23/24] KVM: riscv: selftests: Add a test for counter overflow 本次合入
[PATCH v8 24/24] KVM: riscv: selftests: Add commandline option for SBI PMU test 本次合入

启动host

qemu-system-riscv64 -nographic -cpu rv64,sscofpmf=true,zicond=true,zcb=true,v=true,zvkned=true,zvkg=true,zvkb=true -smp 8 -M virt,pflash0=pflash0,pflash1=pflash1,aia=aplic-imsic,acpi=on,aia-guests=2 -m 4G -serial mon:stdio -blockdev node-name=pflash0,driver=file,read-only=on,filename=RISCV_VIRT_CODE.fd -blockdev node-name=pflash1,driver=file,filename=RISCV_VIRT_VARS.fd -kernel /home/o_zhuzx/rvck/arch/riscv/boot/Image -initrd ./unaligned_cpio/rootfs.new.cpio -append "rdinit=/sbin/init ro console=ttyS0 rootwait earlycon=uart8250"

测试sbi selftest

./sbi_pmu_test 
SBI PMU basic test : PASS
[  672.564941] riscv-kvm-pmu: kvm pmu event creation failed for eidx 1234: -22
SBI PMU event verification test : PASS
SBI PMU event verification with snapshot test : PASS
SBI PMU event verification with overflow test : PASS
# eidx 1234	一个故意传入的无效事件 ID

启动guest

./lkvm run -m 256 -c2 --console serial -p "console=ttyS0 earlycon" --disable-ssaia -k ./Image --debug  --9p /mnt/share,tag=hostshare  --transport mmio

将perf 拷贝到guest

mkdir -p /mnt/share
mount -t 9p "tag=hostshare" /mnt/share -o trans=virtio,version=9p2000.L

先设置低采样率， 防止性能不足host崩溃

echo 100 > /proc/sys/kernel/perf_event_max_sample_rate

>   /mnt/share/perf bench sched messaging -g 5
[  368.068543] perf: interrupt took too long (7325 > 2500), lowering kernel.perf_event_max_sample_rate to 27250
# Running 'sched/messaging' benchmark:
[  368.307203] perf: interrupt took too long (9711 > 9156), lowering kernel.perf_event_max_sample_rate to 20500
[  368.587469] perf: interrupt took too long (12612 > 12138), lowering kernel.perf_event_max_sample_rate to 15750
[  369.082141] perf: interrupt took too long (15911 > 15765), lowering kernel.perf_event_max_sample_rate to 12500
[  369.656568] perf: interrupt took too long (20214 > 19888), lowering kernel.perf_event_max_sample_rate to 9750
[  370.353348] perf: interrupt took too long (25797 > 25267), lowering kernel.perf_event_max_sample_rate to 7750
[  372.899109] perf: interrupt took too long (32535 > 32246), lowering kernel.perf_event_max_sample_rate to 6000
[  376.483151] perf: interrupt took too long (40717 > 40668), lowering kernel.perf_event_max_sample_rate to 4750
[  379.875644] perf: interrupt took too long (51108 > 50896), lowering kernel.perf_event_max_sample_rate to 3750
# 20 sender and receiver processes per group
# 5 groups == 200 processes run

     Total time: 11.397 [sec]
[ perf record: Woken up 1 times to write data ]

[ perf record: Captured and wrote 0.056 MB perf.data (312 samples) ]

# 

report

/mnt/share/perf report --stdio
# To display the perf.data header info, please use --header/--header-only option
#
#
# Total Lost Samples: 0
#
# Samples: 312  of event 'cycles'
# Event count (approx.): 3120000000
#
# Overhead  Command          Shared Object      Symbol                          
# ........  ...............  .................  ................................
#
    16.99%  sched-messaging  [kernel.kallsyms]  [k] finish_task_switch.isra.0
     5.77%  sched-messaging  [kernel.kallsyms]  [k] _raw_spin_unlock_irqrestore
     5.13%  sched-messaging  [kernel.kallsyms]  [k] memset
     4.17%  sched-messaging  [kernel.kallsyms]  [k] handle_softirqs
     3.21%  sched-messaging  [kernel.kallsyms]  [k] __sbi_rfence_v02_call.isra.0
     2.24%  sched-messaging  [kernel.kallsyms]  [k] aa_file_perm
     2.24%  sched-messaging  [kernel.kallsyms]  [k] apparmor_file_permission
     1.92%  sched-messaging  [kernel.kallsyms]  [k] __check_object_size
     1.92%  sched-messaging  [kernel.kallsyms]  [k] set_pte_range
     1.92%  sched-messaging  [kernel.kallsyms]  [k] unix_stream_read_generic
     1.60%  sched-messaging  [kernel.kallsyms]  [k] fallback_scalar_usercopy_sum
     1.60%  sched-messaging  [kernel.kallsyms]  [k] percpu_counter_add_batch
standard input
# 5 groups == 200 processes run

     Total time: 11.397 [sec]
[ perf record: Woken up 1 times to write data ]

record

 /mnt/share/perf record -F 10 -e cycles -- sleep 1
[  252.573692] perf: interrupt took too long (6599 > 2500), lowering kernel.perf_event_max_sample_rate to 30250
[  132.496077] perf: interrupt took too long (19467 > 2500), lowering kernel.perf_event_max_sample_rate to 10250
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.001 MB perf.data (2 samples) ]

report

/mnt/share/perf report --stdio
# To display the perf.data header info, please use --header/--header-only option
#
#
# Total Lost Samples: 0
#
# Samples: 2  of event 'cycles'
# Event count (approx.): 2
#
# Overhead  Command  Shared Object      Symbol                       
# ........  .......  .................  .............................
#
    50.00%  perf-ex  [kernel.kallsyms]  [k] finish_task_switch.isra.0
    50.00%  perf-ex  [kernel.kallsyms]  [k] handle_softirqs

让 KVM 虚拟机（guest）能够高效、准确地使用性能计数器（如 perf 工具），尤其是在支持硬件采样（sampling）的场景下。
#184

[github-actions]

---

开始测试 log: https://github.com/RVCK-Project/rvck/actions/runs/21811438599

参数解析结果

args	value
repository	RVCK-Project/rvck
head ref	pull/218/head
base ref	rvck-6.6
LAVA repo	RVCK-Project/lavaci
LAVA Template	lava-job-template/qemu/qemu-ltp.yaml
Testcase path	lava-testcases/common-test/ltp/ltp.yaml
need run job	kunit-test,kernel-build,check-patch,lava-trigger

---

测试完成

详细结果:

RVCK result

check	result
kunit-test	success
kernel-build	success
lava-trigger	success
check-patch	success

Kunit Test Result

[03:39:40] Testing complete. Ran 457 tests: passed: 445, skipped: 12

Kernel Build Result

Kernel build succeeded: RVCK-Project/rvck/218/

ef4b0a2b488b3fe80c296abb623b20c6 /srv/guix_result/16cc57bf99f215df1c93116cb17c01edeb3cccc3/Image
4736e3403e4120a202590b8f84ccb708 /root/initramfs.img

LAVA Check

args:

• testcase_repo: RVCK-Project/lavaci
• testcase_ref: main
• lava_template: lava-job-template/qemu/qemu-ltp.yaml
• testcase_path: lava-testcases/common-test/ltp/ltp.yaml
• testitem_name: RVCK-Project_rvck_pull_request_target_218__common-test_qemu
• kernel_download_url: https://repo.tarsier-infra.isrc.ac.cn/openEuler-RISC-V/RVCK/OERV-RVCI/RVCK-Project/rvck/218/Image
• initramfs_download_url: https://repo.tarsier-infra.isrc.ac.cn/openEuler-RISC-V/RVCK/OERV-RVCI/RVCK-Project/rvck/218/initramfs.img
• rootfs_download_url: https://fast-mirror.isrc.ac.cn/openeuler-sig-riscv/openEuler-RISC-V/RVCK/openEuler24.03-LTS-SP1/openeuler-rootfs.img.zst

result:

Lava check done! lava log: https://lava.oerv.ac.cn/scheduler/job/1412

lava result count: [fail]: 173, [pass]: 1436, [skip]: 290

Check Patch Result

Total Errors	0
Total Warnings	33

[fangyu0809]

@zhuzhenxxx-collab 在虚拟机里运行perf record的自测试结果，也更新到上面的自测试里面

[zhuzhenxxx-collab]

@zhuzhenxxx-collab 在虚拟机里运行perf record的自测试结果，也更新到上面的自测试里面

Updated

[fangyu0809]

@sterling-teng 自测已完成，帮忙Review合并。

[sterling-teng]

@sterling-teng 自测已完成，帮忙Review合并。

好的。