RELEASE_NOTES

RELEASE NOTES
Intel® System Health Inspector (AKA svr-info)

Fully Supported Platforms
- Xeon Micro-Architectures: GNR,SRF,EMR,SPR,CPX,ICX,CLX,SKX
- Operating Systems: Ubuntu 18.04, 20.04, 22.04, 24.04, CentOS 7, Amazon Linux 2, Debian 11, RHEL 9, Rocky Linux 8
  Note: svr-info may work on other micro-architectures and Linux distributions, but has not been thoroughly tested

2.11.1
Bugs Fixed
- Fix frequency benchmark on GNR
- Add ELC mode for compute dies on SRF and GNR

2.11.0
Features Added
- Report Efficiency Latency Control (ELC) configuration on SRF and GNR
- Report PCIe address, NUMA node, and more for NVME drives
- Add support for PMU profiling on AWS ICX and SPR VMs ... and SRF servers.
- Report L3 size on GNR
- Processor TDP added to Summary field of Excel 'brief' report

Bugs Fixed
- Fix turbo benchmark on systems with many cores, e.g., SRF
- Fix memory performance benchmark when run on systems with large huge page sizes
- Fix Excel report corruption when NaN values result from '-profile pmu'

2.10.0
Features Added
- Support for Sierra Forest Xeon CPUs
- The '-benchmark frequency' option now measures AVX2/3 frequencies in addition to non-AVX frequencies
- Minor improvements to System Summary field in Excel 'Brief' report
- Nodes in multi-node HTML report will now be in the same order as presented in targets file
- SRF and GNR ISAs identified in ISAs report table
- EMR reference benchmark data added

Other Changes
- Removed "Feature" table from report
- Removed Active Idle settings from report

Known Issues
- The '-profile pmu' and '-profile power' options do not function on SRF.
- The storage micro-benchmark may not run on CentOS due to OS locale settings.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population and CPU frequencies may not be accurate on some public cloud IaaS VMs.
 
2.9.1
Bugs Fixed
- Extremely long run time on some EMR virtual machines

2.9.0
Features Added
- Report CPU temperature when CPUs have been active when '-benchmark turbo' is run
- Memory population chart for Dell SPR platform
- Extend prefetchers field to include all known prefetchers

Bugs Fixed
- Slow report generation when an extremely deep call stack is captured with the -analyze option
- LLC size mis-calculated on EMR MCC SKUs
- Micro-architecture mis-identified on AWS m7i instances

Known Issues
- Frequency measurement benchmark not yet supported on GNR and SRF
- The storage micro-benchmark may not run on CentOS due to OS locale settings.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population and CPU frequencies may not be accurate on some public cloud IaaS VMs.

2.8.0
Bugs Fixed
- Turbo on/off incorrectly reported on Excel Brief Report summary field
Features Added
- PMU-based metrics added to profiling options, e.g., -profile pmu or -profile all
- Package and DRAM power stats added to profiling options, e.g., -profile power or -profile all
- Option to skip configuration data collection when profiling or analyzing, e.g., -profile all -noconfig
- Recognize DLB and QAT accelerators on SRF and GNR
- ISA table extended to show availability of newest ISAs
- Add recognition of temperature related entries in System Event Log to Insights
- Recognize GNR, SRF, and AMD Bergamo CPUs
- upgraded to MLC v3.11
Known Issues
- Frequency measurement benchmark not yet supported on GNR and SRF
- The storage micro-benchmark may not run on CentOS due to OS locale settings.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population and CPU frequencies may not be accurate on some public cloud IaaS VMs.

2.7.4
Bugs Fixed
- false positive Vulnerability for CVE-2022-40982 ("Downfall") on some ICX systems

2.7.3
Sub-components Updated
- spectre-meltdown-checker -- updated to latest HEAD commit
- others -- internal components used to collect data and generate reports
Features Added
- None
Bugs Fixed
- None
Known Issues
- The storage micro-benchmark may not run on CentOS due to OS locale settings.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population and CPU frequencies may not be accurate on some public cloud IaaS VMs.

2.7.2
Features Added
- None
Bugs Fixed
- Corrupt Excel reports in svr-info version 2.7.1 caused by incompatibility between Go 1.21 and Excelize package. Reverting to Go 1.20.7.
Known Issues
- The storage micro-benchmark may not run on CentOS due to OS locale settings.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population and CPU frequencies may not be accurate on some public cloud IaaS VMs.

2.7.1
Features Added
- None
Bugs Fixed
- CXL devices were not being detected due to lspci version incompatibility
- Insights that rely on CPU micro-architecture names were not matching appended MCC/XCC text
- Align power-perf policy insight to labels introduced in 2.7.0
Minor Changes
- Use latest (as of release build time) spectre-meltdown-checker script from public repo
- IAX accelerator renamed to IAA
- Fix information text regarding Perl on HTML Analysis report
Known Issues
- The storage micro-benchmark may not run on CentOS due to OS locale settings.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population and CPU frequencies may not be accurate on some public cloud IaaS VMs.

2.7.0
Features Added
- Power & Performance Setting Labels refined. Now includes "Performance", "Balance Performance", "Normal", "Normal Powersave", "Balance Powersave", and "Powersave". Report output also includes specific EPB value, e.g., "Performance (6)".
- Excel 'brief' report's Marketing Claim entry now includes on-CPU accelerator count.
- Extended the Accelerators table to indicate the configuration of work queues for DSA and IAX.
- Extended Insights to trigger if work queues are not configured for DSA and IAX.
Bugs Fixed
- Fixed merging of call stacks collected with the -analyze option.
- Added timeout to collector when attempting to install and uninstall kernel modules to avoid potential hang.
Known Issues
- The storage micro-benchmark may not run on CentOS due to OS locale settings.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population and CPU frequencies may not be accurate on some public cloud IaaS VMs.

2.6.0
Features Added
- Enable MLC on AMD by configuring huge pages
- Improve Summary field format in Excel Brief report
- Display 'Optimized Power Mode' settings in Uncore table
- Add notice to HTML if MLC data missing
Bugs Fixed
- HTML Menu causing content scrolling issues when window re-sized
Known Issues
- The storage micro-benchmark may not run on CentOS due to OS locale settings.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population and CPU frequencies may not be accurate on some public cloud IaaS VMs.

2.5.0
Features Added
- Display file system mount options in Filesystem table
- If file system mounted with 'discard' option, raise a warning in the Insights table
Bugs Fixed
- Calculate and display 'L3 per core' when L3 size is reported in Kilobytes
- Display SPR MCC/XCC microarchitecture from GCP instances
Known Issues
- The storage micro-benchmark may not run on CentOS due to locale settings.
- HTML report may scroll out of view when menu is minimized and window is small.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population may be incorrectly reported on public cloud IaaS VMs.

2.4.0
Features Added
- add DIMM population table for SuperMicro X11DPT-B 
Bugs Fixed
- fix calculation of 'L3 per core' field on multi-socket systems
Known Issues
- The storage micro-benchmark may not run on CentOS due to locale settings.
- HTML report may scroll out of view when menu is minimized and window is small.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population may be incorrectly reported on public cloud IaaS VMs.

2.3.2
Features Added
- report L3 size from MSR, fallback to lscpu
- new field 'L3 per core' in CPU table
- if AMD 'frequency boost' detected, show 'enabled (AMD Frequency Boost)' in the Intel Turbo Boost field
Bugs Fixed
- Don't use "-tt" option with ssh commands that connect to remote sytems
- Fix location of 'extras' directory
- ARM collector binary now extracted for use at runtime
- Graviton 2 CPU architecture detected
Known Issues
- The storage micro-benchmark may not run on CentOS due to locale settings.
- HTML report may scroll out of view when menu is minimized and window is small.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population may be incorrectly reported on public cloud IaaS VMs.

2.3.1
Features Added
- override local temporary directory used by svr-info using the -temp CLI option
  note: override target temporary directory using the -targettemp CLI option (was -temp)
Bugs Fixed
- local temporary directory now created in system's default TEMP location (default)
Known Issues
- The storage micro-benchmark may not run on CentOS due to locale settings.
- HTML report may scroll out of view when menu is minimized and window is small.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population may be incorrectly reported on public cloud IaaS VMs.

2.3.0
Features Added
- Now an all-inclusive single binary (svr-info) for easy re-location
  - to execute the reporter: ./svr-info -reporter "<reporter args>"
  - to execute the collector: ./svr-info -collector "<collector args>"
- Connect to remote machines where keys are configured in ssh-agent
- Version tag added to all report types in new "svr-info" table.
- Differentiate SPR XCC and MCC SKUs
- Support added for Intel Emerald Rapids CPUs
- Recognize Intel Sierra Forest CPUs
- Recognize AMD Genoa CPUs
- Differentiate CPU support vs. Platform support for an ISA
Bugs Fixed
- CXL menu link now navigates to the CXL table
- Remote data collection now stopped when svr-info stopped, e.g., with Ctrl-C
- Output data corruption due to over-aggressive filtering fixed
- DIMM population chart fixed for SuperMicro SPR
- Only include svr-info generated files in output tarball
- Show spec frequencies if measured frequencies not available
- Call calcfreq with correct number of CPUs when some CPUs are off-line
Known Issues
- The storage micro-benchmark may not run on CentOS due to locale settings.
- HTML report may scroll out of view when menu is minimized and window is small.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population may be incorrectly reported on public cloud IaaS VMs.

2.2.0 - First Open Source Release
Features Added
- New/changed command line arguments.
    - "-all" deprecated, use "-benchmark all"
    - "-profile" deprecated, use "-profile all"
    - "-analyze <system|java|all>" added
    - "-cmd_timeout <seconds> added for limiting the time on individual collection commands
    - if in doubt, check the built-in help with the "-h" option
- Recommendations are now called Insights
- Insights expanded to include hi/low CPU utilization, system vulnerabilities, old CPU.
- Insights format now includes recommendation and justification.
- Option added to collect software call stacks for system and Java processes with -analyze system|java|all options.
- Memory profiling option added to Profile option.
- Profiling collection executed before configuration collection for more deterministic timeframe.
- Summary profile table added for profiling data.
- Benchmark Health table renamed to Benchmark Summary and moved to top of Benchmark tab.
- Virtualization feature indicator added to CPU table.
- CPU QDF, if accessible, will be collected by the -megadata data collection option.
- Accelerators table in Configuration now includes full name and description.
- Targets file now supports leading label field and trailing comment, see targets.example file.
Bugs Fixed
- Log error if no targets defined in targets file
- Support passwords that contain '$' in targets file
Known Issues
- The storage micro-benchmark may not run on CentOS due to locale settings.
- HTML report may scroll out of view when menu is minimized and window is small.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population may be incorrectly reported on public cloud IaaS VMs.

2.1.2
Bugs Fixed:
- Correctly identify on-CPU QAT accelerator devices for SPR MCC SKUs. Previous release correctly identified these devices for SPR XCC SKUs but not MCC SKUs.

2.1.1
Bugs Fixed:
- HTML encode (escape) data for insertion into HTML report to limit possibility of script injection security issue

2.1.0
Features Added
- Added perf FlameGraph to Profile report (svr-info -profile).
- JSON report format simplified.
- CXL devices listed in Configuration report.
- PCIe slots listed in Configuration report.
- Uncore table added to Configuration report.
- SPR tunable features listed in Configuration report.
- User Guide now included in installation package.
- Improve usability of charts in HTML report.
- Include Brief report content in JSON report.
Bugs Fixed
- CPU prefetcher settings incorrectly reported
- Incorrectly reporting DDR5 DIMMs as "<OUT OF SPEC>"
Known Issues
- The storage micro-benchmark may not run on CentOS due to locale settings.
- HTML report may scroll out of view when menu is minimized and window is small.
- CPU cache sizes are reported in aggregate on Ubuntu 20.04 and newer.
- DRAM population may be incorrectly reported on public cloud IaaS VMs.

2.0.3
Bugs Fixed
- properly identify SPR Architecture when stepping is 8
- add check for missing 2022 CVEs

2.0.2
Bugs Fixed
- JSON report now includes all data included in HTML and XLSX reports...previously included configuration data only.
- Fix Y-Axis labels on charts in HTML report's profile tab.
- Add missing profile data to XLSX report.
- Do not include loopback NICs or Ubuntu's Snap "loop#" devices in profile data.
- Year was missing from the data field of mktg. claim table in the XLSX brief report.
- Add missing DLB accelerator count field.

2.0.1
Bugs Fixed
- Expand path and file arguments that contain '~' to user's home directory for cases where svr-info is not launched from a shell.
- Fix runtime error that can occur in HTML report generation when some types of PMEM are present in system.

2.0.0
Notable Interface Changes (from 1.x to 2.x)
 - The program name has changed from svr_info to svr-info.
 - The command line arguments can be accessed with a single or double dash, e.g., -cpu or --cpu.
 - HTML, JSON, and XLSX report types are generated by default. Override with the -format command line argument.
 - HTML and XLSX report data is presented on multiple tabs.
 - Combined (all_hosts.*) HTML and XLSX reports will be generated by default when there are two or more hosts specified in the targets file.
 - The JSON report data structure has changed significantly.
 - The CSV report type has been removed.
Features Added
 - System profiling option added (-profile). Collects system performance metrics at configurable frequency (see help).
 - A rules engine is used to offer system configuration recommendations based on current configuration. Rules are configurable.
 - Megadata (--megadata) feature optionally captures additional information in a series of text files. Information collected is configurable.
 - Data from multiple hosts combined into one sheet in combined/multi-host Excel report
 - Excel, HTML and JSON reports now rendered from same data structure for consistency across report types
 - Display versions of Python, Java, and OpenSSL in report
 - Temporary directories now include "svr-info" in their name to more easily identify their origin
 - Full support for Sapphire Rapids - performance reference data, PMU registers, frequency measurement, accelerators
 Bugs Fixed
 - Display full name of CentOS release in report
 - Correctly render DIMM population chart for SPR 8 socket system
 - Do not attempt to resolve hostname during input parameter validation as it may fail under certain circumstances
 - Passwords that include a dollar sign are now supported in targets file if dollar sign proceeded by a backslash
 - If the tar utility is not on the target, display a reasonable error message
Known Issues
- The storage micro-benchmark may not run on CentOS due to locale settings.
- HTML report may scroll out of view when menu is minimized when window is small.

1.4.0
Features Added
- Various report additions and format improvements
- Detect Intel GPUs and report model name
- Report more ISA extensions, e.g. AMX and VAES
- Detection of QAT on platform
- Display PPIN for all installed CPUs
- Ice Lake reference performance data updated
- Sapphire Rapids (D-Step) reference performance data added
- Command line option --input will now search the given directory recursively for .out files
- Align csv/Excel report structure with HTML report
Bugs Fixed
- Can now measure core frequency on Sapphire Rapids
Known Issues
- The storage micro-benchmark may not run on CentOS 8 due to locale settings.
- svr_info will fail to run on a fresh install of CentOS 8 because tar is not installed by default
- passwords that include a dollar sign are not supported for remote target data collection

1.3.0
Features Added
- Support for Sapphire Rapids (no reference performance data)
- Update Ice Lake reference performance data
- Cleanup HTML report layout and add new Power section
- Improved default host data collection performance
Bugs Fixed
- Excel formatted report generation would fail when host names were longer than 32 characters
- Removed micro-benchmarks that measure network metrics due to inconsistency
Known Issues
- The storage micro-benchmark may not run on CentOS 8 due to locale settings.

1.2.4
Features Added
- Recognize additional CPUs for report generation

1.2.3
Bugs Fixed
- Update collector to fix issue with PATH on CentOS 7

1.2.2
Bugs Fixed
- CSV report crash on GCP due to NIC naming scheme in lshw output

1.2.1
Bugs Fixed
- Error when retrieving maximum frequency from dmidecode output

1.2.0
Features Added
- DIMM rank now included in DIMM table
- Maximum Frequency and All-core Maximum Frequency now included in CPU details
Bugs Fixed
- Prefetcher MSRs now retrieved after installing msr kernel driver
- Base Frequency now retrieved from more reliable source
- Data collection now succeeds when PATH is exported in ~/.bashrc on target machine

1.1.0
Features Added
- add support for Ice Lake and Cooper Lake Xeon processors
- add support for Kaby Lake and Coffee Lake Core processors
- reports now include PMU in-use detection
- csv and xlsx reports expanded and reorganized
Bugs Fixed
- crash when output directory is set to current directory using the --output flag
- connection to remote target times out when no user_id is specified
- connection to remote target fails -- retry up to 3 times
- using --ip without --user fails to connect to remote target
- failed to collect requested information if user is 'root'
- crash on AWS EC2 g4dn.metal
- crash if ports used for network performance test are in use

1.0.0
- First Official Release