Sysreport displays an overview of the hardware configuration of the system. It quickly identifies whether or not a hardware feature is available so that you can switch to a different system, if necessary. For example, you might expect a certain hardware feature to be available on a given cloud instance but it isn’t available due to the system configuration.
After running Sysreport, you can examine the output to determine if the configuration is suitable for your needs. Some configuration changes may be required in order to get the system into the desired state for performance analysis. The changes could be as simple as modifying kernel parameters at run-time or more involved, such as recompiling the kernel with different kernel configuration options.
For the AWS EC2 instance on the previous page, Sysreport indicates the global kernel parameter perf_event_paranoid is too strict, meaning that system-level events cannot be monitored by non-privileged users.
Here is the important output:
Actions that can be taken to improve performance tools experience:
...
System-level events can only be monitored by privileged users
set kernel.perf_event_paranoid=0
...
List the kernel parameters and print kernel.perf_event_paranoid by running:
sudo sysctl -a | grep kernel.perf_event_paranoid
The output is:
kernel.perf_event_paranoid = 4
The Linux kernel documentation explains the possible values:
perf_event_paranoid:
Controls use of the performance events system by unprivileged
users (without CAP_SYS_ADMIN). The default value is 2.
-1: Allow use of (almost) all events by all users
Ignore mlock limit after perf_event_mlock_kb without CAP_IPC_LOCK
>=0: Disallow ftrace function tracepoint by users without CAP_SYS_ADMIN
Disallow raw tracepoint access by users without CAP_SYS_ADMIN
>=1: Disallow CPU event access by users without CAP_SYS_ADMIN
>=2: Disallow kernel profiling by users without CAP_SYS_ADMIN
Setting this kernel parameter to 0 allows Linux Perf access to CPU events, kernel profiling, ftrace function tracepoint, and raw tracepoints.
To set the value to 0 until the next reboot run:
sudo sysctl kernel.perf_event_paranoid=0
To permanently set the value, use a text editor to add the following line to the file /etc/sysctl.conf
kernel.perf_event_paranoid=0
Running Sysreport again displays the new configuration:
System feature report:
Collected: 2024-03-21 14:12:53.529899
Script version: 2024-03-21 11:50:23.254919
Running as root: False
System hardware:
Architecture: ARMv8.4
CPUs: 16
CPU types: 16 x ARM Neoverse V1 r1p1
cache info: size, associativity, sharing
cache line size: 64
Caches:
16 x L1D 64K 4-way 64b-line
16 x L1I 64K 4-way 64b-line
16 x L2U 1M 8-way 64b-line
1 x L3U 32M 16-way 64b-line
System memory: 30.8G
Atomic operations: True
interconnect: <unknown> x 1
NUMA nodes: 1
Sockets: 1
OS configuration:
Kernel: 6.5.0
config: /boot/config-6.5.0-1015-aws
32-bit support: True
build dir: /lib/modules/6.5.0-1015-aws/build
uses atomics: True
huge pages: False
transparent: madvise
MPAM configured: False
resctrl: False
Distribution: Ubuntu 22.04.3 LTS
libc version: glibc 2.35
boot info: ACPI
KPTI enabled: False
Lockdown: landlock, lockdown, yama, integrity, apparmor
Mitigations: spectre_v2:CSV2, BHB; spec_store_bypass:Speculative Store Bypass disabled via prctl; spectre_v1:__user pointer sanitization
Performance features:
perf tools: True
perf installed at: /usr/lib/linux-tools/6.5.0-1015-aws/perf
perf with OpenCSD: False
perf counters: 2
perf sampling: None
perf HW trace: None
perf paranoid: 0
perf in userspace: disabled
interconnect perf: None
/proc/kcore: True
/dev/mem: True
Actions that can be taken to improve performance tools experience:
perf tools cannot decode hardware trace
build with CORESIGHT=1
non-invasive sampling (SPE) not enabled
ensure APIC table describes SPE interrupt
kernel module arm_spe_pmu.ko must be built
hardware trace not enabled
rebuild kernel with CONFIG_CORESIGHT
ensure ACPI describes CoreSight trace fabric