selftests/resctrl: Add resctrl selftest
Subsystems (2)
v7 cover selftests/resctrl: Add resctrl selftest
With more and more resctrl features are being added by Intel, AMD and ARM, a test tool is becoming more and more useful to validate that both hardware and software functionalities work as expected.
We introduce resctrl selftest to cover resctrl features on both X86 and ARM architectures. It first implements MBM (Memory Bandwidth Monitoring), MBA (Memory Bandwidth Allocation), L3 CAT (Cache Allocation Technology), and CQM (Cache QoS Monitoring) tests. We can enhance the selftest tool to include more functionality tests in future.
The tool has been tested on both Intel RDT and AMD QoS.
There is an existing resctrl test suite 'intel_cmt_cat'. But the major purpose of the tool is to test Intel(R) RDT hardware via writing and reading MSR registers. It does access resctrl file system; but the functionalities are very limited. And it doesn't support automatic test and a lot of manual verifications are involved.
So the selftest tool we are introducing here provides a convenient tool which does automatic resctrl testing, is easily available in kernel tree, and will be extended to AMD QoS and ARM MPAM.
The selftest tool is in tools/testing/selftests/resctrl in order to have generic test code for all architectures.
Changelog:
v7:
- Fix a few warnings when compiling patches separately, pointed by Babu
v6:
- Fix a benchmark reading optimized out issue in newer GCC.
- Fix a few coding style issues.
- Re-arrange code among patches to make cleaner code. No change in patches structure.
v5:
- Based the v4 patches submitted by Fenghua Yu and added changes to support AMD.
- Changed the function name get_sock_num to get_resource_id. Intel uses socket number for schemata and AMD uses l3 index id. To generalize, changed the function name to get_resource_id.
- Added the code to detect vendor.
- Disabled the few tests for AMD where the test results are not clear. Also AMD does not have IMC.
- Fixed few compile issues.
- Some cleanup to make each patch independent.
- Tested the patches on AMD system. Fenghua, Need your help to test on Intel box. Please feel free to change and resubmit if something broken.
- Here is the link for previous version.
https://lore.kernel.org/lkml/1545438038-75107-1-git-send-email-fenghua.yu=
@intel.com/
v4:
- address comments from Balu and Randy
- Add CAT and CQM tests
v3:
- Change code based on comments from Babu Moger
- Remove some unnessary code and use pipe to communicate b/w processes
v2:
- Change code based on comments from Babu Moger
- Clean up other places.
Arshiya Hayatkhan Pathan (4):
selftests/resctrl: Add MBM test
selftests/resctrl: Add MBA test
selftests/resctrl: Add Cache QoS Monitoring (CQM) selftest selftests/resctrl: Add Cache Allocation Technology (CAT) selftest
Babu Moger (3):
selftests/resctrl: Add vendor detection mechanism selftests/resctrl: Use cache index3 id for AMD schemata masks selftests/resctrl: Disable MBA and MBM tests for AMD
Fenghua Yu (2):
selftests/resctrl: Add README for resctrl tests
selftests/resctrl: Add the test in MAINTAINERS
Sai Praneeth Prakhya (4):
selftests/resctrl: Add basic resctrl file system operations and data selftests/resctrl: Read memory bandwidth from perf IMC counter and
from resctrl file system
selftests/resctrl: Add callback to start a benchmark selftests/resctrl: Add built in benchmark
MAINTAINERS | 1 + tools/testing/selftests/resctrl/Makefile | 16 + tools/testing/selftests/resctrl/README | 54 ++ tools/testing/selftests/resctrl/cache.c | 274 +++++++++ tools/testing/selftests/resctrl/cat_test.c | 242 ++++++++ tools/testing/selftests/resctrl/cqm_test.c | 173 ++++++ tools/testing/selftests/resctrl/fill_buf.c | 210 +++++++ tools/testing/selftests/resctrl/mba_test.c | 178 ++++++ tools/testing/selftests/resctrl/mbm_test.c | 150 +++++ tools/testing/selftests/resctrl/resctrl.h | 117 ++++ tools/testing/selftests/resctrl/resctrl_tests.c | 238 ++++++++ tools/testing/selftests/resctrl/resctrl_val.c | 723 ++++++++++++++++++++++++ tools/testing/selftests/resctrl/resctrlfs.c | 668 ++++++++++++++++++++++ 13 files changed, 3044 insertions(+) create mode 100644 tools/testing/selftests/resctrl/Makefile create mode 100644 tools/testing/selftests/resctrl/README create mode 100644 tools/testing/selftests/resctrl/cache.c create mode 100644 tools/testing/selftests/resctrl/cat_test.c create mode 100644 tools/testing/selftests/resctrl/cqm_test.c create mode 100644 tools/testing/selftests/resctrl/fill_buf.c create mode 100644 tools/testing/selftests/resctrl/mba_test.c create mode 100644 tools/testing/selftests/resctrl/mbm_test.c create mode 100644 tools/testing/selftests/resctrl/resctrl.h create mode 100644 tools/testing/selftests/resctrl/resctrl_tests.c create mode 100644 tools/testing/selftests/resctrl/resctrl_val.c create mode 100644 tools/testing/selftests/resctrl/resctrlfs.c
Signature · 1 line
2.7.4
PATCH v7 1/13 selftests/resctrl: Add README for resctrl tests
resctrl tests will be implemented. README is added for the tool first.
--- tools/testing/selftests/resctrl/README | 54 ++++++++++++++++++++++++++++++++++ 1 file changed, 54 insertions(+) create mode 100644 tools/testing/selftests/resctrl/README
tools/testing/selftests/resctrl/README new file +54 -0
@@ -0,0 +1,54 @@ +resctrl_tests - resctrl file system test suit + +Authors: + Fenghua Yu <[email protected]> + Sai Praneeth Prakhya <[email protected]>, + Arshiya Hayatkhan Pathan <[email protected]> + +resctrl_tests tests various resctrl functionalities and interfaces including +both software and hardware. + +Currently it supports Memory Bandwidth Monitoring test and Memory Bandwidth +Allocation test on Intel RDT hardware. More tests will be added in the future. +And the test suit can be extended to cover AMD QoS and ARM MPAM hardware +as well. + +BUILD +----- + +Run "make" to build executable file "resctrl_tests". + +RUN +--- + +To use resctrl_tests, root or sudoer privileges are required. This is because +the test needs to mount resctrl file system and change contents in the file +system. + +Executing the test without any parameter will run all supported tests: + + sudo ./resctrl_tests + +OVERVIEW OF EXECUTION +--------------------- + +A test case has four stages: + + - setup: mount resctrl file system, create group, setup schemata, move test + process pids to tasks, start benchmark. + - execute: let benchmark run + - verify: get resctrl data and verify the data with another source, e.g. + perf event. + - teardown: umount resctrl and clear temporary files. + +ARGUMENTS +--------- + +Parameter '-h' shows usage information. + +usage: resctrl_tests [-h] [-b "benchmark_cmd [options]"] [-t test list] [-n no_of_bits] + -b benchmark_cmd [options]: run specified benchmark for MBM, MBA and CQM default benchmark is builtin fill_buf + -t test list: run tests specified in the test list, e.g. -t mbm, mba, cqm, cat + -n no_of_bits: run cache tests using specified no of bits in cache bit mask + -p cpu_no: specify CPU number to run the test. 1 is default + -h: help
Signature · 1 line
2.7.4
PATCH v7 2/13 selftests/resctrl: Add basic resctrl file system operations and data
In reply to Fenghua YuFrom: Sai Praneeth Prakhya <[email protected]>
The basic resctrl file system operations and data are added for future usage by resctrl selftest tool.
--- tools/testing/selftests/resctrl/Makefile | 10 + tools/testing/selftests/resctrl/resctrl.h | 48 +++ tools/testing/selftests/resctrl/resctrlfs.c | 464 ++++++++++++++++++++++++++++ 3 files changed, 522 insertions(+) create mode 100644 tools/testing/selftests/resctrl/Makefile create mode 100644 tools/testing/selftests/resctrl/resctrl.h create mode 100644 tools/testing/selftests/resctrl/resctrlfs.c
tools/testing/selftests/resctrl/Makefile new file +10 -0
@@ -0,0 +1,10 @@ +CC = gcc +CFLAGS = -g -Wall + +*.o: *.c + $(CC) $(CFLAGS) -c *.c + +.PHONY: clean + +clean: + $(RM) *.o *~
tools/testing/selftests/resctrl/resctrl.h new file +48 -0
48 lines not on this page · raw message
tools/testing/selftests/resctrl/resctrlfs.c new file +464 -0
464 lines not on this page · raw message
Signature · 1 line
2.7.4
Re: [PATCH v7 02/13] selftests/resctrl: Add basic resctrl file system operations and data
On Sat, 9 Feb 2019 18:50:31 -0800
Fenghua Yu <[email protected]> wrote:
Hi,
some comments inline.
Quoted 24 lines
From: Sai Praneeth Prakhya <[email protected]>
The basic resctrl file system operations and data are added for future usage by resctrl selftest tool.
Signed-off-by Sai Praneeth Prakhya <[email protected]>Signed-off-by Arshiya Hayatkhan Pathan <[email protected]>Signed-off-by Fenghua Yu <[email protected]>Signed-off-by Babu Moger <[email protected]>--- tools/testing/selftests/resctrl/Makefile | 10 + tools/testing/selftests/resctrl/resctrl.h | 48 +++ tools/testing/selftests/resctrl/resctrlfs.c | 464 ++++++++++++++++++++++++++++ 3 files changed, 522 insertions(+) create mode 100644 tools/testing/selftests/resctrl/Makefile create mode 100644 tools/testing/selftests/resctrl/resctrl.h create mode 100644 tools/testing/selftests/resctrl/resctrlfs.ctools/testing/selftests/resctrl/Makefile new file +0
-0
tools/testing/selftests/resctrl/Makefile line 1
Quoted 1 line
tools/testing/selftests/resctrl/Makefile new file +1
-0@@ -0,0 +1,10 @@ +CC = gcc
Changing this to
CC = $(CROSS_COMPILE)gcc
make this cross compileable.
tools/testing/selftests/resctrl/Makefile line 2
Quoted 1 line
tools/testing/selftests/resctrl/Makefile new file +1
-0+CFLAGS = -g -Wall
Can we add -O here? For once -O0 generates horrible code, but also -O tends to catch more bugs.
tools/testing/selftests/resctrl/Makefile line 5
Quoted 3 lines
tools/testing/selftests/resctrl/Makefile new file +3
-0+ +*.o: *.c + $(CC) $(CFLAGS) -c *.c
This is a built-in rule in make, so you can remove it here:
https://www.gnu.org/software/make/manual/html_node/Catalogue-of-Rules.html#Catalogue-of-Rules
Quoted 128 lines
tools/testing/selftests/resctrl/Makefile new file +5
-0+ +.PHONY: clean + +clean: + $(RM) *.o *~tools/testing/selftests/resctrl/resctrl.h new file +48
-0@@ -0,0 +1,48 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#define _GNU_SOURCE +#ifndef RESCTRL_H +#define RESCTRL_H +#include <stdio.h> +#include <errno.h> +#include <sched.h> +#include <stdlib.h> +#include <unistd.h> +#include <string.h> +#include <signal.h> +#include <dirent.h> +#include <stdbool.h> +#include <sys/stat.h> +#include <sys/ioctl.h> +#include <sys/mount.h> +#include <sys/types.h> +#include <asm/unistd.h> +#include <linux/perf_event.h> + +#define RESCTRL_PATH "/sys/fs/resctrl" +#define PHYS_ID_PATH "/sys/devices/system/cpu/cpu" + +#define PARENT_EXIT (err_msg) \ + do { \ + perror(err_msg); \ + kill(ppid, SIGKILL); \ + exit(EXIT_FAILURE); \ + } while (0) + +pid_t bm_pid, ppid; + +int remount_resctrlfs (bool mum_resctrlfs); +int get_resource_id (int cpu_no,int *resource_id); +int validate_bw_report_request (char *bw_report); +int validate_resctrl_feature_request (char *resctrl_val); +int taskset_benchmark (pid_t bm_pid,int cpu_no); +void run_benchmark (int signum,siginfo_t *info,void *ucontext); +int write_schemata (char *ctrlgrp,char *schemata,int cpu_no, +char *resctrl_val); +int write_bm_pid_to_resctrl (pid_t bm_pid,char *ctrlgrp,char *mongrp, +char *resctrl_val); +int perf_event_open (struct perf_event_attr *hw_event,pid_t pid,int cpu, +int group_fd,unsigned long flags); +int run_fill_buf (unsigned long span,int malloc_and_init_memory,int memflush, +int op,char *resctrl_va); + +#endif /* RESCTRL_H */ tools/testing/selftests/resctrl/resctrlfs.c new file +63
-0@@ -0,0 +1,464 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Basic resctrl file system operations +* +* Copyright (C) 2018 Intel Corporation +* +* Authors: +* Arshiya Hayatkhan Pathan <[email protected]> +* Sai Praneeth Prakhya <[email protected]>, +* Fenghua Yu <[email protected]> +*/ +#include "resctrl.h" + +#define RESCTRL_MBM "L3 monitoring detected" +#define RESCTRL_MBA "MB allocation detected" +enum { +RESCTRL_FEATURE_MBM , +RESCTRL_FEATURE_MBA , +MAX_RESCTRL_FEATURES +}; + +/* +* remount_resctrlfs - Remount resctrl FS at /sys/fs/resctrl +* @mum_resctrlfs: Should the resctrl FS be remounted? +* +* If not mounted, mount it. +* If mounted and mum_resctrlfs then remount resctrl FS. +* If mounted and !mum_resctrlfs then noop +* +* Return: 0 on success, non-zero on failure +*/ +int remount_resctrlfs (bool mum_resctrlfs) +{ +DIR *dp; +struct dirent *ep; +unsigned int count =0 ; + +/* +* If kernel is built with CONFIG_RESCTRL, then /sys/fs/resctrl should +* be present by default +*/ + dp =opendir (RESCTRL_PATH ); +if (dp) { +while ((ep =readdir (dp)) !=NULL ) + count++; +closedir (dp); + }else { +perror ("Unable to read /sys/fs/resctrl" ); + +return -1 ; + } + +/* +* If resctrl FS has more than two entries, it means that resctrl FS has +* already been mounted. The two default entries are "." and "..", these +* are present even when resctrl FS is not mounted +*/ +if (count >2 ) { +if (mum_resctrlfs) { +if (umount (RESCTRL_PATH ) !=0 ) { +perror ("Unable to umount resctrl" ); + +return errno;
tools/testing/selftests/resctrl/resctrlfs.c line 73
Quoted 10 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +10
-0+ } +printf ("Remount: done!\n" ); + }else { +printf ("Mounted already. Not remounting!\n" ); + +return 0 ; + } + } + +if (mount ("resctrl" ,RESCTRL_PATH ,"resctrl" ,0 ,NULL ) !=0 ) {
Don't we need to consider mount options at some point? According to Documentation/x86/resctrl_ui.txt there is cdp, cdpl2 and mba_MBps, which we need to set to get certain features.
Quoted 322 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +322
-0+perror ("Unable to mount resctrl FS at /sys/fs/resctrl" ); + +return errno; + } + +return 0 ; +} + +int umount_resctrlfs (void ) +{ +if (umount (RESCTRL_PATH )) { +perror ("Unable to umount resctrl" ); + +return errno; + } + +return 0 ; +} + +/* +* get_resource_id - Get socket number/l3 id for a specified CPU +* @cpu_no: CPU number +* @resource_id: Socket number or l3_id +* +* Return: >= 0 on success, < 0 on failure. +*/ +int get_resource_id (int cpu_no,int *resource_id) +{ +char phys_pkg_path[1024 ]; +FILE *fp; + +sprintf (phys_pkg_path,"%s%d/topology/physical_package_id" , +PHYS_ID_PATH , cpu_no); + fp =fopen (phys_pkg_path,"r" ); +if (!fp) { +perror ("Failed to open physical_package_id" ); + +return -1 ; + } +if (fscanf (fp,"%d" , resource_id) <=0 ) { +perror ("Could not get socket number or l3 id" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + +return 0 ; +} + +/* +* taskset_benchmark - Taskset PID (i.e. benchmark) to a specified cpu +* @bm_pid: PID that should be binded +* @cpu_no: CPU number at which the PID would be binded +* +* Return: 0 on success, non-zero on failure +*/ +int taskset_benchmark (pid_t bm_pid,int cpu_no) +{ +cpu_set_t my_set; + +CPU_ZERO (&my_set); +CPU_SET (cpu_no, &my_set); + +if (sched_setaffinity (bm_pid,sizeof (cpu_set_t), &my_set)) { +perror ("Unable to taskset benchmark" ); + +return -1 ; + } + +printf ("Taskset benchmark: done!\n" ); + +return 0 ; +} + +/* +* run_benchmark - Run a specified benchmark or fill_buf (default benchmark) +* in specified signal. Direct benchmark stdio to /dev/null. +* @signum: signal number +* @info: signal info +* @ucontext: user context in signal handling +* +* Return: void +*/ +void run_benchmark (int signum,siginfo_t *info,void *ucontext) +{ +unsigned long long span; +int operation, ret; +char **benchmark_cmd; +FILE *fp; + + benchmark_cmd = info->si_ptr; + +/* +* Direct stdio of child to /dev/null, so that only parent writes to +* stdio (console) +*/ + fp =freopen ("/dev/null" ,"w" , stdout); +if (!fp) +PARENT_EXIT ("Unable to direct benchmark status to /dev/null" ); + +if (strcmp (benchmark_cmd[0 ],"fill_buf" ) ==0 ) { +/* Execute default fill_buf benchmark */ + span =strtoul (benchmark_cmd[1 ],NULL ,10 ); + operation =atoi (benchmark_cmd[4 ]); +if (run_fill_buf (span,1 ,1 , operation,NULL )) +fprintf (stderr,"Error in running fill buffer\n" ); + }else { +/* Execute specified benchmark */ + ret =execvp (benchmark_cmd[0 ], benchmark_cmd); +if (ret) +perror ("wrong\n" ); + } + +fclose (stdout); +PARENT_EXIT ("Unable to run specified benchmark" ); +} + +/* +* create_grp - Create a group only if one doesn't exist +* @grp_name: Name of the group +* @grp: Full path and name of the group +* @parent_grp: Full path and name of the parent group +* +* Return: 0 on success, non-zero on failure +*/ +static int create_grp (const char *grp_name,char *grp,const char *parent_grp) +{ +int found_grp =0 ; +struct dirent *ep; +DIR *dp; + +/* Check if requested grp exists or not */ + dp =opendir (parent_grp); +if (dp) { +while ((ep =readdir (dp)) !=NULL ) { +if (strcmp (ep->d_name, grp_name) ==0 ) + found_grp =1 ; + } +closedir (dp); + }else { +perror ("Unable to open resctrl for group" ); + +return -1 ; + } + +/* Requested grp doesn't exist, hence create it */ +if (found_grp ==0 ) { +if (mkdir (grp,0 ) ==-1 ) { +perror ("Unable to create group" ); + +return -1 ; + } + } + +return 0 ; +} + +static int write_pid_to_tasks (char *tasks,pid_t pid) +{ +FILE *fp; + + fp =fopen (tasks,"w" ); +if (!fp) { +perror ("Failed to open tasks file" ); + +return -1 ; + } +if (fprintf (fp,"%d\n" , pid) <0 ) { +perror ("Failed to wr pid to tasks file" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + +return 0 ; +} + +/* +* write_bm_pid_to_resctrl - Write a PID (i.e. benchmark) to resctrl FS +* @bm_pid: PID that should be written +* @ctrlgrp: Name of the control monitor group (con_mon grp) +* @mongrp: Name of the monitor group (mon grp) +* @resctrl_val: Resctrl feature (Eg: mbm, mba.. etc) +* +* If a con_mon grp is requested, create it and write pid to it, otherwise +* write pid to root con_mon grp. +* If a mon grp is requested, create it and write pid to it, otherwise +* pid is not written, this means that pid is in con_mon grp and hence +* should consult con_mon grp's mon_data directory for results. +* +* Return: 0 on success, non-zero on failure +*/ +int write_bm_pid_to_resctrl (pid_t bm_pid,char *ctrlgrp,char *mongrp, +char *resctrl_val) +{ +char controlgroup[256 ], monitorgroup[256 ], monitorgroup_p[256 ]; +char tasks[256 ]; +int ret; + +if (ctrlgrp) +sprintf (controlgroup,"%s/%s" ,RESCTRL_PATH , ctrlgrp); +else +sprintf (controlgroup,"%s" ,RESCTRL_PATH ); + +/* Create control and monitoring group and write pid into it */ + ret =create_grp (ctrlgrp, controlgroup,RESCTRL_PATH ); +if (ret) +return ret; +sprintf (tasks,"%s/tasks" , controlgroup); + ret =write_pid_to_tasks (tasks, bm_pid); +if (ret) +return ret; + +/* Create mon grp and write pid into it for "mbm" test */ +if ((strcmp (resctrl_val,"mbm" ) ==0 )) { +if (mongrp) { +sprintf (monitorgroup_p,"%s/mon_groups" , controlgroup); +sprintf (monitorgroup,"%s/%s" , monitorgroup_p, mongrp); + ret =create_grp (mongrp, monitorgroup, monitorgroup_p); +if (ret) +return ret; + +sprintf (tasks,"%s/mon_groups/%s/tasks" , + controlgroup, mongrp); + ret =write_pid_to_tasks (tasks, bm_pid); +if (ret) +return ret; + } + } + +printf ("Write benchmark to resctrl FS: done!\n" ); + +return 0 ; +} + +/* +* write_schemata - Update schemata of a con_mon grp +* @ctrlgrp: Name of the con_mon grp +* @schemata: Schemata that should be updated to +* @cpu_no: CPU number that the benchmark PID is binded to +* @resctrl_val: Resctrl feature (Eg: mbm, mba.. etc) +* +* Update schemata of a con_mon grp *only* if requested resctrl feature is +* allocation type +* +* Return: 0 on success, non-zero on failure +*/ +int write_schemata (char *ctrlgrp,char *schemata,int cpu_no,char *resctrl_val) +{ +char controlgroup[1024 ], schema[1024 ]; +int resource_id; +FILE *fp; + +if (strcmp (resctrl_val,"mba" ) ==0 ) { +if (!schemata) { +printf ("Schemata empty, so not updating\n" ); + +return 0 ; + } +if (get_resource_id (cpu_no, &resource_id) <0 ) { +perror ("Failed to get resource id" ); +return -1 ; + } + +if (ctrlgrp) +sprintf (controlgroup,"%s/%s/schemata" ,RESCTRL_PATH , + ctrlgrp); +else +sprintf (controlgroup,"%s/schemata" ,RESCTRL_PATH ); +sprintf (schema,"%s%d%c%s" ,"MB:" , resource_id,'=' , schemata); + + fp =fopen (controlgroup,"w" ); +if (!fp) { +perror ("Failed to open control group" ); + +return -1 ; + } + +if (fprintf (fp,"%s\n" , schema) <0 ) { +perror ("Failed to write schemata in control group" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + +printf ("Write schemata to resctrl FS: done!\n" ); + } + +return 0 ; +} + +/* +* validate_resctrl_feature_request - Check if requested feature is valid. +* @resctrl_val: Requested feature +* +* Return: 0 on success, non-zero on failure +*/ +int validate_resctrl_feature_request (char *resctrl_val) +{ +int resctrl_features_supported[MAX_RESCTRL_FEATURES ]; +const char *resctrl_features_list[MAX_RESCTRL_FEATURES ] = { +"mbm" ,"mba" }; +int i, valid_resctrl_feature =-1 ; +char line[1024 ]; +FILE *fp; + +if (!resctrl_val) { +fprintf (stderr,"resctrl feature cannot be NULL\n" ); + +return -1 ; + } + +for (i =0 ; i <MAX_RESCTRL_FEATURES ; i++) + resctrl_features_supported[i] =0 ; + +/* Is the resctrl feature request valid? */ +for (i =0 ; i <MAX_RESCTRL_FEATURES ; i++) { +if (strcmp (resctrl_features_list[i], resctrl_val) ==0 ) + valid_resctrl_feature = i;
tools/testing/selftests/resctrl/resctrlfs.c line 405
Quoted 10 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +10
-0+ } +if (valid_resctrl_feature ==-1 ) { +fprintf (stderr,"Not a valid resctrl feature request\n" ); + +return -1 ; + } + +/* Enumerate resctrl features supported by this platform */ +if (system ("dmesg > dmesg" ) !=0 ) { +perror ("Could not create custom dmesg file" );
This fails horribly if the local directory is not writable. Creating a pipe (similar to how we communicate between the benchmark processes) and reading from there avoid this.
Quoted 4 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +4
-0+ +return -1 ; + } +
tools/testing/selftests/resctrl/resctrlfs.c line 419
Quoted 10 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +10
-0+ fp =fopen ("dmesg" ,"r" ); +if (!fp) { +perror ("Could not read custom created dmesg" ); + +return -1 ; + } + +while (fgets (line,1024 , fp)) { +if ((strstr (line,RESCTRL_MBM )) !=NULL ) + resctrl_features_supported[RESCTRL_FEATURE_MBM ] =1 ;
In general this approach is not very reliable, as the beginning of the kernel log could have been overwritten already in the dmesg buffer. Also this is not the way we should detect features: I think the content of /sys/fs/resctrl/info should be used for that purpose.
Cheers,
Andre.
Quoted 45 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +45
-0+if ((strstr (line,RESCTRL_MBA )) !=NULL ) + resctrl_features_supported[RESCTRL_FEATURE_MBA ] =1 ; + } +fclose (fp); + +if (system ("rm -rf dmesg" ) !=0 ) +perror ("Unable to remove 'dmesg' file" ); + +/* Is the resctrl feature request supported? */ +if (!resctrl_features_supported[valid_resctrl_feature]) { +fprintf (stderr,"resctrl feature not supported!" ); + +return -1 ; + } + +return 0 ; +} + +int validate_bw_report_request (char *bw_report) +{ +if (strcmp (bw_report,"reads" ) ==0 ) +return 0 ; +if (strcmp (bw_report,"writes" ) ==0 ) +return 0 ; +if (strcmp (bw_report,"nt-writes" ) ==0 ) { +strcpy (bw_report,"writes" ); +return 0 ; + } +if (strcmp (bw_report,"total" ) ==0 ) +return 0 ; + +fprintf (stderr,"Requested iMC B/W report type unavailable\n" ); + +return -1 ; +} + +int perf_event_open (struct perf_event_attr *hw_event,pid_t pid,int cpu, +int group_fd,unsigned long flags) +{ +int ret; + + ret =syscall (__NR_perf_event_open, hw_event, pid, cpu, + group_fd, flags); +return ret; +}
RE: [PATCH v7 02/13] selftests/resctrl: Add basic resctrl file system operations and data
Quoted 41 lines
From: Andre Przywara [mailto:[email protected]] On Sat, 9 Feb 2019 18:50:31 -0800
Fenghua Yu <[email protected]> wrote:Hi,
some comments inline.
Quoted 27 lines
From: Sai Praneeth Prakhya <[email protected]>
The basic resctrl file system operations and data are added for future usage by resctrl selftest tool.
Signed-off-by Sai Praneeth Prakhya <[email protected]>Signed-off-by Arshiya Hayatkhan PathanSigned-off-by Fenghua Yu <[email protected]>Signed-off-by Babu Moger <[email protected]>--- tools/testing/selftests/resctrl/Makefile | 10 + tools/testing/selftests/resctrl/resctrl.h | 48 +++ tools/testing/selftests/resctrl/resctrlfs.c | 464 ++++++++++++++++++++++++++++ 3 files changed, 522 insertions(+) create mode 100644 tools/testing/selftests/resctrl/Makefile create mode 100644 tools/testing/selftests/resctrl/resctrl.h create mode 100644 tools/testing/selftests/resctrl/resctrlfs.ctools/testing/selftests/resctrl/Makefile new file +0
-0tools/testing/selftests/resctrl/Makefile line 1
Quoted 1 line
tools/testing/selftests/resctrl/Makefile new file +1
-0@@ -0,0 +1,10 @@ +CC = gccChanging this to
CC = $(CROSS_COMPILE)gcc
make this cross compileable.
Quoted 4 lines
tools/testing/selftests/resctrl/Makefile line 2
Quoted 1 line
tools/testing/selftests/resctrl/Makefile new file +1
-0+CFLAGS = -g -WallCan we add -O here? For once -O0 generates horrible code, but also -O tends to catch more bugs.
Sure.
Quoted 8 lines
tools/testing/selftests/resctrl/Makefile line 5
Quoted 4 lines
tools/testing/selftests/resctrl/Makefile new file +3
-0+ +*.o: *.c + $(CC) $(CFLAGS) -c *.cThis is a built-in rule in make, so you can remove it here:
https://www.gnu.org/software/make/manual/html_node/Catalogue-of-
Rules.html#Catalogue-of-Rules
Will do.
Quoted 126 lines
Quoted 33 lines
tools/testing/selftests/resctrl/Makefile new file +5
-0+ +.PHONY: clean + +clean: + $(RM) *.o *~tools/testing/selftests/resctrl/resctrl.h new file +20
-0@@ -0,0 +1,48 @@ +/* SPDX-License-Identifier: GPL-2.0 */ #define _GNU_SOURCE #ifndef +RESCTRL_H #define RESCTRL_H #include <stdio.h> #include <errno.h> +#include <sched.h> #include <stdlib.h> #include <unistd.h> #include +<string.h> #include <signal.h> #include <dirent.h> #include +<stdbool.h> #include <sys/stat.h> #include <sys/ioctl.h> #include +<sys/mount.h> #include <sys/types.h> #include <asm/unistd.h> #include +<linux/perf_event.h> + +#define RESCTRL_PATH "/sys/fs/resctrl" +#define PHYS_ID_PATH "/sys/devices/system/cpu/cpu" + +#define PARENT_EXIT (err_msg) \ + do { \ + perror(err_msg); \ + kill(ppid, SIGKILL); \ + exit(EXIT_FAILURE); \ + } while (0) + +pid_t bm_pid, ppid; +tools/testing/selftests/resctrl/resctrl.h line 30
Quoted 10 lines
tools/testing/selftests/resctrl/resctrl.h new file +10
-0+int remount_resctrlfs (bool mum_resctrlfs);int get_resource_id (int +cpu_no,int *resource_id);int validate_bw_report_request (char +*bw_report);int validate_resctrl_feature_request (char *resctrl_val); +int taskset_benchmark (pid_t bm_pid,int cpu_no);void +run_benchmark (int signum,siginfo_t *info,void *ucontext);int +write_schemata (char *ctrlgrp,char *schemata,int cpu_no, +char *resctrl_val); +int write_bm_pid_to_resctrl (pid_t bm_pid,char *ctrlgrp,char *mongrp, +char *resctrl_val); +int perf_event_open(struct perf_event_attr *hw_event, pid_t pid, intcpu,
tools/testing/selftests/resctrl/resctrl.h line 31
Quoted 1 line
tools/testing/selftests/resctrl/resctrl.h new file +1
-0+ int group_fd, unsigned long flags); intrun_fill_buf(unsigned
Quoted 53 lines
tools/testing/selftests/resctrl/resctrl.h new file +4
-0+long span, int malloc_and_init_memory, int memflush, + int op, char *resctrl_va); + +#endif /* RESCTRL_H */ tools/testing/selftests/resctrl/resctrlfs.c new file +42
-0@@ -0,0 +1,464 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Basic resctrl file system operations +* +* Copyright (C) 2018 Intel Corporation +* +* Authors: +* Arshiya Hayatkhan Pathan <[email protected]> +* Sai Praneeth Prakhya <[email protected]>, +* Fenghua Yu <[email protected]> +*/ +#include "resctrl.h" + +#define RESCTRL_MBM "L3 monitoring detected" +#define RESCTRL_MBA "MB allocation detected" +enum { +RESCTRL_FEATURE_MBM , +RESCTRL_FEATURE_MBA , +MAX_RESCTRL_FEATURES +}; + +/* +* remount_resctrlfs - Remount resctrl FS at /sys/fs/resctrl +* @mum_resctrlfs: Should the resctrl FS be remounted? +* +* If not mounted, mount it. +* If mounted and mum_resctrlfs then remount resctrl FS. +* If mounted and !mum_resctrlfs then noop +* +* Return: 0 on success, non-zero on failure */ int +remount_resctrlfs (bool mum_resctrlfs) { +DIR *dp; +struct dirent *ep; +unsigned int count =0 ; + +/* +* If kernel is built with CONFIG_RESCTRL, then /sys/fs/resctrl should +* be present by default +*/ + dp =opendir (RESCTRL_PATH ); +if (dp) { +while ((ep =readdir (dp)) !=NULL )tools/testing/selftests/resctrl/resctrlfs.c line 52
Quoted 10 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +10
-0+ count++; +closedir (dp); + }else { +perror ("Unable to read /sys/fs/resctrl" ); + +return -1 ; + } + +/* +* If resctrl FS has more than two entries, it means that resctrl FS has
tools/testing/selftests/resctrl/resctrlfs.c line 53
Quoted 1 line
tools/testing/selftests/resctrl/resctrlfs.c new file +1
-0+* already been mounted. The two default entries are "." and "..", these
Quoted 14 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +14
-0+* are present even when resctrl FS is not mounted +*/ +if (count >2 ) { +if (mum_resctrlfs) { +if (umount (RESCTRL_PATH ) !=0 ) { +perror ("Unable to umount resctrl" ); + +return errno; + } +printf ("Remount: done!\n" ); + }else { +printf ("Mounted already. Not remounting!\n" ); + +return 0 ;
Quoted 8 lines
tools/testing/selftests/resctrl/resctrlfs.c line 73
Quoted 4 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +4
-0+ } + } + +if (mount ("resctrl" ,RESCTRL_PATH ,"resctrl" ,0 ,NULL ) !=0 ) {Don't we need to consider mount options at some point? According to
Documentation/x86/resctrl_ui.txtthere is cdp, cdpl2 andmba_MBps, which we need to set to get certain features.
As said in the cover patch, there more tests in the future to cover cdp, mba, etc. We will use mount options in those future tests.
It's good to mount without any parameter for the tests in the current patch set.
Quoted 355 lines
Quoted 65 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +64
-0+perror ("Unable to mount resctrl FS at /sys/fs/resctrl" ); + +return errno; + } + +return 0 ; +} + +int umount_resctrlfs (void ) +{ +if (umount (RESCTRL_PATH )) { +perror ("Unable to umount resctrl" ); + +return errno; + } + +return 0 ; +} + +/* +* get_resource_id - Get socket number/l3 id for a specified CPU +* @cpu_no: CPU number +* @resource_id: Socket number or l3_id +* +* Return: >= 0 on success, < 0 on failure. +*/ +int get_resource_id (int cpu_no,int *resource_id) { +char phys_pkg_path[1024 ]; +FILE *fp; + +sprintf (phys_pkg_path,"%s%d/topology/physical_package_id" , +PHYS_ID_PATH , cpu_no); + fp =fopen (phys_pkg_path,"r" ); +if (!fp) { +perror ("Failed to open physical_package_id" ); + +return -1 ; + } +if (fscanf (fp,"%d" , resource_id) <=0 ) { +perror ("Could not get socket number or l3 id" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + +return 0 ; +} + +/* +* taskset_benchmark - Taskset PID (i.e. benchmark) to a specified cpu +* @bm_pid: PID that should be binded +* @cpu_no: CPU number at which the PID would be binded +* +* Return: 0 on success, non-zero on failure */ int +taskset_benchmark (pid_t bm_pid,int cpu_no) { +cpu_set_t my_set; + +CPU_ZERO (&my_set); +CPU_SET (cpu_no, &my_set); + +if (sched_setaffinity (bm_pid,sizeof (cpu_set_t), &my_set)) { +perror ("Unable to taskset benchmark" ); +tools/testing/selftests/resctrl/resctrlfs.c
Quoted 10 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +10
-0+return -1 ; + } + +printf ("Taskset benchmark: done!\n" ); + +return 0 ; +} + +/* +* run_benchmark - Run a specified benchmark or fill_buf (default benchmark)
Quoted 12 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +12
-0+* in specified signal. Direct benchmark stdio to /dev/null. +* @signum: signal number +* @info: signal info +* @ucontext: user context in signal handling +* +* Return: void +*/ +void run_benchmark (int signum,siginfo_t *info,void *ucontext) { +unsigned long long span; +int operation, ret; +char **benchmark_cmd; +FILE *fp;tools/testing/selftests/resctrl/resctrlfs.c
Quoted 10 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +10
-0+ + benchmark_cmd = info->si_ptr; + +/* +* Direct stdio of child to /dev/null, so that only parent writes to +* stdio (console) +*/ + fp =freopen ("/dev/null" ,"w" , stdout); +if (!fp) + PARENT_EXIT("Unable to direct benchmark status to
/dev/null");Quoted 107 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +107
-0+ +if (strcmp (benchmark_cmd[0 ],"fill_buf" ) ==0 ) { +/* Execute default fill_buf benchmark */ + span =strtoul (benchmark_cmd[1 ],NULL ,10 ); + operation =atoi (benchmark_cmd[4 ]); +if (run_fill_buf (span,1 ,1 , operation,NULL )) +fprintf (stderr,"Error in running fill buffer\n" ); + }else { +/* Execute specified benchmark */ + ret =execvp (benchmark_cmd[0 ], benchmark_cmd); +if (ret) +perror ("wrong\n" ); + } + +fclose (stdout); +PARENT_EXIT ("Unable to run specified benchmark" ); } + +/* +* create_grp - Create a group only if one doesn't exist +* @grp_name: Name of the group +* @grp: Full path and name of the group +* @parent_grp: Full path and name of the parent group +* +* Return: 0 on success, non-zero on failure */ static int +create_grp (const char *grp_name,char *grp,const char *parent_grp) { +int found_grp =0 ; +struct dirent *ep; +DIR *dp; + +/* Check if requested grp exists or not */ + dp =opendir (parent_grp); +if (dp) { +while ((ep =readdir (dp)) !=NULL ) { +if (strcmp (ep->d_name, grp_name) ==0 ) + found_grp =1 ; + } +closedir (dp); + }else { +perror ("Unable to open resctrl for group" ); + +return -1 ; + } + +/* Requested grp doesn't exist, hence create it */ +if (found_grp ==0 ) { +if (mkdir (grp,0 ) ==-1 ) { +perror ("Unable to create group" ); + +return -1 ; + } + } + +return 0 ; +} + +static int write_pid_to_tasks (char *tasks,pid_t pid) { +FILE *fp; + + fp =fopen (tasks,"w" ); +if (!fp) { +perror ("Failed to open tasks file" ); + +return -1 ; + } +if (fprintf (fp,"%d\n" , pid) <0 ) { +perror ("Failed to wr pid to tasks file" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + +return 0 ; +} + +/* +* write_bm_pid_to_resctrl - Write a PID (i.e. benchmark) to resctrl FS +* @bm_pid: PID that should be written +* @ctrlgrp: Name of the control monitor group (con_mon grp) +* @mongrp: Name of the monitor group (mon grp) +* @resctrl_val: Resctrl feature (Eg: mbm, mba.. etc) +* +* If a con_mon grp is requested, create it and write pid to it, +otherwise +* write pid to root con_mon grp. +* If a mon grp is requested, create it and write pid to it, +otherwise +* pid is not written, this means that pid is in con_mon grp and +hence +* should consult con_mon grp's mon_data directory for results. +* +* Return: 0 on success, non-zero on failure */ int +write_bm_pid_to_resctrl (pid_t bm_pid,char *ctrlgrp,char *mongrp, +char *resctrl_val) +{ +char controlgroup[256 ], monitorgroup[256 ], monitorgroup_p[256 ]; +char tasks[256 ]; +int ret; + +if (ctrlgrp) +sprintf (controlgroup,"%s/%s" ,RESCTRL_PATH , ctrlgrp); +else +sprintf (controlgroup,"%s" ,RESCTRL_PATH ); + +/* Create control and monitoring group and write pid into it */ + ret =create_grp (ctrlgrp, controlgroup,RESCTRL_PATH ); +if (ret)tools/testing/selftests/resctrl/resctrlfs.c
Quoted 10 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +10
-0+return ret; +sprintf (tasks,"%s/tasks" , controlgroup); + ret =write_pid_to_tasks (tasks, bm_pid); +if (ret) +return ret; + +/* Create mon grp and write pid into it for "mbm" test */ +if ((strcmp (resctrl_val,"mbm" ) ==0 )) { +if (mongrp) { + sprintf(monitorgroup_p, "%s/mon_groups",controlgroup);
tools/testing/selftests/resctrl/resctrlfs.c line 285
Quoted 1 line
tools/testing/selftests/resctrl/resctrlfs.c new file +1
-0+ sprintf(monitorgroup, "%s/%s", monitorgroup_p,mongrp);
tools/testing/selftests/resctrl/resctrlfs.c line 286
Quoted 1 line
tools/testing/selftests/resctrl/resctrlfs.c new file +1
-0+ ret = create_grp(mongrp, monitorgroup,
monitorgroup_p);Quoted 11 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +11
-0+if (ret) +return ret; + +sprintf (tasks,"%s/mon_groups/%s/tasks" , + controlgroup, mongrp); + ret =write_pid_to_tasks (tasks, bm_pid); +if (ret) +return ret; + } + } +tools/testing/selftests/resctrl/resctrlfs.c
Quoted 10 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +10
-0+printf ("Write benchmark to resctrl FS: done!\n" ); + +return 0 ; +} + +/* +* write_schemata - Update schemata of a con_mon grp +* @ctrlgrp: Name of the con_mon grp +* @schemata: Schemata that should be updated to +* @cpu_no: CPU number that the benchmark PID is binded to
Quoted 16 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +16
-0+* @resctrl_val: Resctrl feature (Eg: mbm, mba.. etc) +* +* Update schemata of a con_mon grp *only* if requested resctrl +feature is +* allocation type +* +* Return: 0 on success, non-zero on failure */ int +write_schemata (char *ctrlgrp,char *schemata,int cpu_no,char +*resctrl_val) { +char controlgroup[1024 ], schema[1024 ]; +int resource_id; +FILE *fp; + +if (strcmp (resctrl_val,"mba" ) ==0 ) { +if (!schemata) { +printf ("Schemata empty, so not updating\n" );tools/testing/selftests/resctrl/resctrlfs.c
Quoted 10 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +10
-0+ +return 0 ; + } +if (get_resource_id (cpu_no, &resource_id) <0 ) { +perror ("Failed to get resource id" ); +return -1 ; + } + +if (ctrlgrp) + sprintf(controlgroup, "%s/%s/schemata",
RESCTRL_PATH,tools/testing/selftests/resctrl/resctrlfs.c
Quoted 3 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +3
-0+ ctrlgrp); + else + sprintf(controlgroup, "%s/schemata",
RESCTRL_PATH);tools/testing/selftests/resctrl/resctrlfs.c line 337
Quoted 1 line
tools/testing/selftests/resctrl/resctrlfs.c new file +1
-0+ sprintf(schema, "%s%d%c%s", "MB:", resource_id, '=',schemata);
Quoted 48 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +48
-0+ + fp =fopen (controlgroup,"w" ); +if (!fp) { +perror ("Failed to open control group" ); + +return -1 ; + } + +if (fprintf (fp,"%s\n" , schema) <0 ) { +perror ("Failed to write schemata in control group" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + +printf ("Write schemata to resctrl FS: done!\n" ); + } + +return 0 ; +} + +/* +* validate_resctrl_feature_request - Check if requested feature is valid. +* @resctrl_val: Requested feature +* +* Return: 0 on success, non-zero on failure */ int +validate_resctrl_feature_request (char *resctrl_val) { +int resctrl_features_supported[MAX_RESCTRL_FEATURES ]; +const char *resctrl_features_list[MAX_RESCTRL_FEATURES ] = { +"mbm" ,"mba" }; +int i, valid_resctrl_feature =-1 ; +char line[1024 ]; +FILE *fp; + +if (!resctrl_val) { +fprintf (stderr,"resctrl feature cannot be NULL\n" ); + +return -1 ; + } + +for (i =0 ; i <MAX_RESCTRL_FEATURES ; i++) + resctrl_features_supported[i] =0 ; + +/* Is the resctrl feature request valid? */ +for (i =0 ; i <MAX_RESCTRL_FEATURES ; i++) { +if (strcmp (resctrl_features_list[i], resctrl_val) ==0 ) + valid_resctrl_feature = i;tools/testing/selftests/resctrl/resctrlfs.c line 405
Quoted 10 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +10
-0+ } +if (valid_resctrl_feature ==-1 ) { +fprintf (stderr,"Not a valid resctrl feature request\n" ); + +return -1 ; + } + +/* Enumerate resctrl features supported by this platform */ +if (system ("dmesg > dmesg" ) !=0 ) { +perror ("Could not create custom dmesg file" );This fails horribly if the local directory is not writable. Creating a pipe (similar to how we communicate between the benchmark processes) and reading from there avoid this.
Quoted 4 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +4
-0+ +return -1 ; + } +tools/testing/selftests/resctrl/resctrlfs.c line 424
Quoted 10 lines
tools/testing/selftests/resctrl/resctrlfs.c new file +10
-0+ fp =fopen ("dmesg" ,"r" ); +if (!fp) { +perror ("Could not read custom created dmesg" ); + +return -1 ; + } + +while (fgets (line,1024 , fp)) { +if ((strstr (line,RESCTRL_MBM )) !=NULL ) +resctrl_features_supported[RESCTRL_FEATURE_MBM] = 1;
Quoted 4 lines
In general this approach is not very reliable, as the beginning of the kernel log could have been overwritten already in the dmesg buffer.
Also this is not the way we should detect features: I think the content of/sys/fs/resctrl/infoshould be used for that purpose.
Agree with you. We will use /sys/fs/resctrl/info directory to enumerate features instead of dmesg.
Thanks.
-Fenghua
Re: [PATCH v7 02/13] selftests/resctrl: Add basic resctrl file system operations and data
In reply to Fenghua YuQuoted 9 lines · Fenghua Yu
From: Sai Praneeth Prakhya <[email protected]>
The basic resctrl file system operations and data are added for future usage by resctrl selftest tool.
Signed-off-by Sai Praneeth Prakhya <[email protected]>Signed-off-by Arshiya Hayatkhan Pathan <[email protected]>Signed-off-by Fenghua Yu <[email protected]>Signed-off-by Babu Moger <[email protected]>
And while feedback from Andre is being addressed, you those SOB chains need to be fixed.
Thx.
Signature · 4 lines
Regards/Gruss,
Boris.
Good mailing practices for 400: avoid top-posting and trim the reply.RE: [PATCH v7 02/13] selftests/resctrl: Add basic resctrl file system operations and data
Quoted 7 lines
From: Borislav Petkov [mailto:[email protected]]
Quoted 5 lines · Fenghua Yu
From: Sai Praneeth Prakhya <[email protected]>
The basic resctrl file system operations and data are added for future usage by resctrl selftest tool.
Quoted 8 lines
Quoted 5 lines
Signed-off-by Sai Praneeth Prakhya <[email protected]>Signed-off-by Arshiya Hayatkhan PathanSigned-off-by Fenghua Yu <[email protected]>Signed-off-by Babu Moger <[email protected]>And while feedback from Andre is being addressed, you those SOB chains need to be fixed.
Will fix the SOB chains.
Thanks.
-Fenghua
PATCH v7 3/13 selftests/resctrl: Read memory bandwidth from perf IMC counter and from resctrl file system
In reply to Fenghua YuFrom: Sai Praneeth Prakhya <[email protected]>
Total memory bandwidth can be monitored from perf IMC counter and from resctrl file system. Later the two will be compared to verify the total memory bandwidth read from resctrl is correct.
--- tools/testing/selftests/resctrl/resctrl_val.c | 118 ++++++++++++++++++++++++++ 1 file changed, 118 insertions(+) create mode 100644 tools/testing/selftests/resctrl/resctrl_val.c
tools/testing/selftests/resctrl/resctrl_val.c new file +118 -0
@@ -0,0 +1,118 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Memory bandwidth monitoring and allocation library +* +* Copyright (C) 2018 Intel Corporation +* +* Authors: +* Arshiya Hayatkhan Pathan <[email protected]> +* Sai Praneeth Prakhya <[email protected]>, +* Fenghua Yu <[email protected]> +*/ +#include "resctrl.h" + +#define UNCORE_IMC "uncore_imc" +#define READ_FILE_NAME "events/cas_count_read" +#define WRITE_FILE_NAME "events/cas_count_write" +#define DYN_PMU_PATH "/sys/bus/event_source/devices" +#define SCALE 0.00006103515625 +#define MAX_IMCS 20 +#define MAX_TOKENS 5 +#define READ 0 +#define WRITE 1 +#define CON_MON_MBM_LOCAL_BYTES_PATH \ + "%s/%s/mon_groups/%s/mon_data/mon_L3_%02d/mbm_local_bytes" + +#define CON_MBM_LOCAL_BYTES_PATH \ + "%s/%s/mon_data/mon_L3_%02d/mbm_local_bytes" + +#define MON_MBM_LOCAL_BYTES_PATH \ + "%s/mon_groups/%s/mon_data/mon_L3_%02d/mbm_local_bytes" + +#define MBM_LOCAL_BYTES_PATH \ + "%s/mon_data/mon_L3_%02d/mbm_local_bytes" + +struct membw_read_format { +__u64 value;/* The value of the event */ +__u64 time_enabled;/* if PERF_FORMAT_TOTAL_TIME_ENABLED */ +__u64 time_running;/* if PERF_FORMAT_TOTAL_TIME_RUNNING */ +__u64 id;/* if PERF_FORMAT_ID */ +}; + +struct imc_counter_config { +__u32 type; +__u64 event; +__u64 umask; +struct perf_event_attr pe; +struct membw_read_format return_value; +int fd; +}; + +static struct imc_counter_config imc_counters_config[MAX_IMCS ][2 ]; + +void membw_initialize_perf_event_attr (int i,int j) +{ +memset (&imc_counters_config[i][j].pe,0 , +sizeof (struct perf_event_attr )); + imc_counters_config[i][j].pe.type = imc_counters_config[i][j].type; + imc_counters_config[i][j].pe.size =sizeof (struct perf_event_attr ); + imc_counters_config[i][j].pe.disabled =1 ; + imc_counters_config[i][j].pe.inherit =1 ; + imc_counters_config[i][j].pe.exclude_guest =1 ; + imc_counters_config[i][j].pe.config = + imc_counters_config[i][j].umask <<8 | + imc_counters_config[i][j].event; + imc_counters_config[i][j].pe.sample_type =PERF_SAMPLE_IDENTIFIER ; + imc_counters_config[i][j].pe.read_format = +PERF_FORMAT_TOTAL_TIME_ENABLED |PERF_FORMAT_TOTAL_TIME_RUNNING ; +} + +void membw_ioctl_perf_event_ioc_reset_enable (int i,int j) +{ +ioctl (imc_counters_config[i][j].fd,PERF_EVENT_IOC_RESET ,0 ); +ioctl (imc_counters_config[i][j].fd,PERF_EVENT_IOC_ENABLE ,0 ); +} + +void membw_ioctl_perf_event_ioc_disable (int i,int j) +{ +ioctl (imc_counters_config[i][j].fd,PERF_EVENT_IOC_DISABLE ,0 ); +} + +/* +* get_event_and_umask: Parse config into event and umask +* @cas_count_cfg: Config +* @count: iMC number +* @op: Operation (read/write) +*/ +void get_event_and_umask (char *cas_count_cfg,int count,bool op) +{ +char *token[MAX_TOKENS ]; +int i =0 ; + +strcat (cas_count_cfg,"," ); + token[0 ] =strtok (cas_count_cfg,"=," ); + +for (i =1 ; i <MAX_TOKENS ; i++) + token[i] =strtok (NULL ,"=," ); + +for (i =0 ; i <MAX_TOKENS ; i++) { +if (!token[i]) +break ; +if (strcmp (token[i],"event" ) ==0 ) { +if (op ==READ ) + imc_counters_config[count][READ ].event = +strtol (token[i +1 ],NULL ,16 ); +else + imc_counters_config[count][WRITE ].event = +strtol (token[i +1 ],NULL ,16 ); + } +if (strcmp (token[i],"umask" ) ==0 ) { +if (op ==READ ) + imc_counters_config[count][READ ].umask = +strtol (token[i +1 ],NULL ,16 ); +else + imc_counters_config[count][WRITE ].umask = +strtol (token[i +1 ],NULL ,16 ); + } + } +}
Signature · 1 line
2.7.4
PATCH v7 4/13 selftests/resctrl: Add callback to start a benchmark
In reply to Fenghua YuFrom: Sai Praneeth Prakhya <[email protected]>
The callback starts a child process and puts the child pid in created resctrl group with specified memory bandwidth in schemata. The child starts running benchmark.
--- tools/testing/selftests/resctrl/resctrl.h | 27 ++ tools/testing/selftests/resctrl/resctrl_val.c | 582 ++++++++++++++++++++++++++ 2 files changed, 609 insertions(+)
tools/testing/selftests/resctrl/resctrl.h +27 -0
@@ -3,6 +3,7 @@#ifndef RESCTRL_H #define RESCTRL_H #include <stdio.h> +#include <stdarg.h> #include <errno.h> #include <sched.h> #include <stdlib.h>
@@ -28,10 +29,35 @@exit (EXIT_FAILURE ); \ }while (0 ) +/* +* resctrl_val_param: resctrl test parameters +* @resctrl_val: Resctrl feature (Eg: mbm, mba.. etc) +* @ctrlgrp: Name of the control monitor group (con_mon grp) +* @mongrp: Name of the monitor group (mon grp) +* @cpu_no: CPU number to which the benchmark would be binded +* @span: Memory bytes accessed in each benchmark iteration +* @mum_resctrlfs: Should the resctrl FS be remounted? +* @filename: Name of file to which the o/p should be written +* @bw_report: Bandwidth report type (reads vs writes) +* @setup: Call back function to setup test environment +*/ +struct resctrl_val_param { +char *resctrl_val; +char ctrlgrp[64 ]; +char mongrp[64 ]; +int cpu_no; +int span; +int mum_resctrlfs; +char filename[64 ]; +char *bw_report; +int (*setup)(int num, ...); +}; +pid_t bm_pid, ppid;int remount_resctrlfs (bool mum_resctrlfs);int get_resource_id (int cpu_no,int *resource_id); +int umount_resctrlfs (void );int validate_bw_report_request (char *bw_report);int validate_resctrl_feature_request (char *resctrl_val);int taskset_benchmark (pid_t bm_pid,int cpu_no);
@@ -44,5 +70,6 @@ int perf_event_open(struct perf_event_attr *hw_event, pid_t pid, int cpu, int group_fd, unsigned long flags);int run_fill_buf (unsigned long span,int malloc_and_init_memory,int memflush,int op,char *resctrl_va); +int resctrl_val (char **benchmark_cmd,struct resctrl_val_param *param);#endif /* RESCTRL_H */
tools/testing/selftests/resctrl/resctrl_val.c +582 -0
597 lines not on this page · raw message
Signature · 1 line
2.7.4
Re: [PATCH v7 04/13] selftests/resctrl: Add callback to start a benchmark
On Sat, 9 Feb 2019 18:50:33 -0800
Fenghua Yu <[email protected]> wrote:
Hi,
Quoted 185 lines
From: Sai Praneeth Prakhya <[email protected]>
The callback starts a child process and puts the child pid in created resctrl group with specified memory bandwidth in schemata. The child starts running benchmark.
Signed-off-by Sai Praneeth Prakhya <[email protected]>Signed-off-by Arshiya Hayatkhan Pathan <[email protected]>Signed-off-by Fenghua Yu <[email protected]>Signed-off-by Babu Moger <[email protected]>--- tools/testing/selftests/resctrl/resctrl.h | 27 ++ tools/testing/selftests/resctrl/resctrl_val.c | 582 ++++++++++++++++++++++++++ 2 files changed, 609 insertions(+)tools/testing/selftests/resctrl/resctrl.h +27
-0@@ -3,6 +3,7 @@#ifndef RESCTRL_H #define RESCTRL_H #include <stdio.h> +#include <stdarg.h> #include <errno.h> #include <sched.h> #include <stdlib.h> @@ -28,10 +29,35 @@exit (EXIT_FAILURE ); \ }while (0 ) +/* +* resctrl_val_param: resctrl test parameters +* @resctrl_val: Resctrl feature (Eg: mbm, mba.. etc) +* @ctrlgrp: Name of the control monitor group (con_mon grp) +* @mongrp: Name of the monitor group (mon grp) +* @cpu_no: CPU number to which the benchmark would be binded +* @span: Memory bytes accessed in each benchmark iteration +* @mum_resctrlfs: Should the resctrl FS be remounted? +* @filename: Name of file to which the o/p should be written +* @bw_report: Bandwidth report type (reads vs writes) +* @setup: Call back function to setup test environment +*/ +struct resctrl_val_param { +char *resctrl_val; +char ctrlgrp[64 ]; +char mongrp[64 ]; +int cpu_no; +int span; +int mum_resctrlfs; +char filename[64 ]; +char *bw_report; +int (*setup)(int num, ...); +}; +pid_t bm_pid, ppid;int remount_resctrlfs (bool mum_resctrlfs);int get_resource_id (int cpu_no,int *resource_id); +int umount_resctrlfs (void );int validate_bw_report_request (char *bw_report);int validate_resctrl_feature_request (char *resctrl_val);int taskset_benchmark (pid_t bm_pid,int cpu_no);@@ -44,5 +70,6 @@ int perf_event_open(struct perf_event_attr *hw_event, pid_t pid, int cpu, int group_fd, unsigned long flags);int run_fill_buf (unsigned long span,int malloc_and_init_memory,int memflush,int op,char *resctrl_va); +int resctrl_val (char **benchmark_cmd,struct resctrl_val_param *param);#endif /* RESCTRL_H */ tools/testing/selftests/resctrl/resctrl_val.c +93
-0@@ -11,6 +11,7 @@*/ #include "resctrl.h" +#define MB (1024 * 1024)#define UNCORE_IMC "uncore_imc"#define READ_FILE_NAME "events/cas_count_read"#define WRITE_FILE_NAME "events/cas_count_write"@@ -48,6 +49,8 @@ struct imc_counter_config {int fd; }; +static char mbm_total_path[1024 ]; +static int imcs;static struct imc_counter_config imc_counters_config[MAX_IMCS ][2 ];void membw_initialize_perf_event_attr (int i,int j)@@ -116,3 +119,582 @@ void get_event_and_umask(char *cas_count_cfg, int count, bool op) } } } + +static int open_perf_event (int i,int cpu_no,int j) +{ + imc_counters_config[i][j].fd = +perf_event_open (&imc_counters_config[i][j].pe,-1 , cpu_no,-1 , +PERF_FLAG_FD_CLOEXEC ); + +if (imc_counters_config[i][j].fd ==-1 ) { +fprintf (stderr,"Error opening leader %llx\n" , + imc_counters_config[i][j].pe.config); + +return -1 ; + } + +return 0 ; +} + +/* Get type and config (read and write) of an iMC counter */ +static int read_from_imc_dir (char *imc_dir,int count) +{ +char cas_count_cfg[1024 ], imc_counter_cfg[1024 ], imc_counter_type[1024 ]; +FILE *fp; + +/* Get type of iMC counter */ +sprintf (imc_counter_type,"%s%s" , imc_dir,"type" ); + fp =fopen (imc_counter_type,"r" ); +if (!fp) { +perror ("Failed to open imc counter type file" ); + +return -1 ; + } +if (fscanf (fp,"%u" , &imc_counters_config[count][READ ].type) <=0 ) { +perror ("Could not get imc type" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + + imc_counters_config[count][WRITE ].type = + imc_counters_config[count][READ ].type; + +/* Get read config */ +sprintf (imc_counter_cfg,"%s%s" , imc_dir,READ_FILE_NAME ); + fp =fopen (imc_counter_cfg,"r" ); +if (!fp) { +perror ("Failed to open imc config file" ); + +return -1 ; + } +if (fscanf (fp,"%s" , cas_count_cfg) <=0 ) { +perror ("Could not get imc cas count read" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + +get_event_and_umask (cas_count_cfg, count,READ ); + +/* Get write config */ +sprintf (imc_counter_cfg,"%s%s" , imc_dir,WRITE_FILE_NAME ); + fp =fopen (imc_counter_cfg,"r" ); +if (!fp) { +perror ("Failed to open imc config file" ); + +return -1 ; + } +if (fscanf (fp,"%s" , cas_count_cfg) <=0 ) { +perror ("Could not get imc cas count write" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + +get_event_and_umask (cas_count_cfg, count,WRITE ); + +return 0 ; +} + +/* +* A system can have 'n' number of iMC (Integrated Memory Controller) +* counters, get that 'n'. For each iMC counter get it's type and config. +* Also, each counter has two configs, one for read and the other for write. +* A config again has two parts, event and umask. +* Enumerate all these details into an array of structures. +* +* Return: >= 0 on success. < 0 on failure. +*/
tools/testing/selftests/resctrl/resctrl_val.c void get_event_and_umask(char *cas_count_cfg, int count, bool op) · line 221
Quoted 10 lines
tools/testing/selftests/resctrl/resctrl_val.c +10
-0+static int num_of_imcs (void ) +{ +unsigned int count =0 ; +char imc_dir[1024 ]; +struct dirent *ep; +int ret; +DIR *dp; + + dp =opendir (DYN_PMU_PATH ); +if (dp) {
Changing this to: if (!dp) {
and handling the error case first makes this more readable and avoids the indentation.
Quoted 226 lines
tools/testing/selftests/resctrl/resctrl_val.c +226
-0+while ((ep =readdir (dp))) { +if (strstr (ep->d_name,UNCORE_IMC )) { +sprintf (imc_dir,"%s/%s/" ,DYN_PMU_PATH , + ep->d_name); + ret =read_from_imc_dir (imc_dir, count); +if (ret) { +closedir (dp); + +return ret; + } + count++; + } + } +closedir (dp); +if (count ==0 ) { +perror ("Unable find iMC counters!\n" ); + +return -1 ; + } + }else { +perror ("Unable to open PMU directory!\n" ); + +return -1 ; + } + +return count; +} + +static int initialize_mem_bw_imc (void ) +{ +int imc, j; + + imcs =num_of_imcs (); +if (imcs <=0 ) +return imcs; + +/* Initialize perf_event_attr structures for all iMC's */ +for (imc =0 ; imc < imcs; imc++) { +for (j =0 ; j <2 ; j++) +membw_initialize_perf_event_attr (imc, j); + } + +return 0 ; +} + +/* +* get_mem_bw_imc: Memory band width as reported by iMC counters +* @cpu_no: CPU number that the benchmark PID is binded to +* @bw_report: Bandwidth report type (reads, writes) +* +* Memory B/W utilized by a process on a socket can be calculated using +* iMC counters. Perf events are used to read these counters. +* +* Return: >= 0 on success. < 0 on failure. +*/ +static float get_mem_bw_imc (int cpu_no,char *bw_report) +{ +float reads, writes, of_mul_read, of_mul_write; +int imc, j, ret; + +/* Start all iMC counters to log values (both read and write) */ + reads =0 , writes =0 , of_mul_read =1 , of_mul_write =1 ; +for (imc =0 ; imc < imcs; imc++) { +for (j =0 ; j <2 ; j++) { + ret =open_perf_event (imc, cpu_no, j); +if (ret) +return -1 ; + } +for (j =0 ; j <2 ; j++) +membw_ioctl_perf_event_ioc_reset_enable (imc, j); + } + +sleep (1 ); + +/* Stop counters after a second to get results (both read and write) */ +for (imc =0 ; imc < imcs; imc++) { +for (j =0 ; j <2 ; j++) +membw_ioctl_perf_event_ioc_disable (imc, j); + } + +/* +* Get results which are stored in struct type imc_counter_config +* Take over flow into consideration before calculating total b/w +*/ +for (imc =0 ; imc < imcs; imc++) { +struct imc_counter_config *r = + &imc_counters_config[imc][READ ]; +struct imc_counter_config *w = + &imc_counters_config[imc][WRITE ]; + +if (read (r->fd, &r->return_value, +sizeof (struct membw_read_format )) ==-1 ) { +perror ("Couldn't get read b/w through iMC" ); + +return -1 ; + } + +if (read (w->fd, &w->return_value, +sizeof (struct membw_read_format )) ==-1 ) { +perror ("Couldn't get write bw through iMC" ); + +return -1 ; + } + +__u64 r_time_enabled = r->return_value.time_enabled; +__u64 r_time_running = r->return_value.time_running; + +if (r_time_enabled != r_time_running) + of_mul_read = (float )r_time_enabled / + (float )r_time_running; + +__u64 w_time_enabled = w->return_value.time_enabled; +__u64 w_time_running = w->return_value.time_running; + +if (w_time_enabled != w_time_running) + of_mul_write = (float )w_time_enabled / + (float )w_time_running; + reads += r->return_value.value * of_mul_read *SCALE ; + writes += w->return_value.value * of_mul_write *SCALE ; + } + +for (imc =0 ; imc < imcs; imc++) { +close (imc_counters_config[imc][READ ].fd); +close (imc_counters_config[imc][WRITE ].fd); + } + +if (strcmp (bw_report,"reads" ) ==0 ) +return reads; + +if (strcmp (bw_report,"writes" ) ==0 ) +return writes; + +return (reads + writes); +} + +void set_mbm_path (const char *ctrlgrp,const char *mongrp,int resource_id) +{ +if (ctrlgrp && mongrp) +sprintf (mbm_total_path,CON_MON_MBM_LOCAL_BYTES_PATH , +RESCTRL_PATH , ctrlgrp, mongrp, resource_id); +else if (!ctrlgrp && mongrp) +sprintf (mbm_total_path,MON_MBM_LOCAL_BYTES_PATH ,RESCTRL_PATH , + mongrp, resource_id); +else if (ctrlgrp && !mongrp) +sprintf (mbm_total_path,CON_MBM_LOCAL_BYTES_PATH ,RESCTRL_PATH , + ctrlgrp, resource_id); +else if (!ctrlgrp && !mongrp) +sprintf (mbm_total_path,MBM_LOCAL_BYTES_PATH ,RESCTRL_PATH , + resource_id); +} + +/* +* initialize_mem_bw_resctrl: Appropriately populate "mbm_total_path" +* @ctrlgrp: Name of the control monitor group (con_mon grp) +* @mongrp: Name of the monitor group (mon grp) +* @cpu_no: CPU number that the benchmark PID is binded to +* @resctrl_val: Resctrl feature (Eg: mbm, mba.. etc) +*/ +static void initialize_mem_bw_resctrl (const char *ctrlgrp,const char *mongrp, +int cpu_no,char *resctrl_val) +{ +int resource_id; + +if (get_resource_id (cpu_no, &resource_id) <0 ) { +perror ("Could not get resource_id" ); +return ; + } + +if (strcmp (resctrl_val,"mbm" ) ==0 ) +set_mbm_path (ctrlgrp, mongrp, resource_id); + +if ((strcmp (resctrl_val,"mba" ) ==0 )) { +if (ctrlgrp) +sprintf (mbm_total_path,CON_MBM_LOCAL_BYTES_PATH , +RESCTRL_PATH , ctrlgrp, resource_id); +else +sprintf (mbm_total_path,MBM_LOCAL_BYTES_PATH , +RESCTRL_PATH , resource_id); + } +} + +/* +* Get MBM Local bytes as reported by resctrl FS +* For MBM, +* 1. If con_mon grp and mon grp are given, then read from con_mon grp's mon grp +* 2. If only con_mon grp is given, then read from con_mon grp +* 3. If both are not given, then read from root con_mon grp +* For MBA, +* 1. If con_mon grp is given, then read from it +* 2. If con_mon grp is not given, then read from root con_mon grp +*/ +static unsigned long get_mem_bw_resctrl (void ) +{ +unsigned long mbm_total =0 ; +FILE *fp; + + fp =fopen (mbm_total_path,"r" ); +if (!fp) { +perror ("Failed to open total bw file" ); + +return -1 ; + } +if (fscanf (fp,"%lu" , &mbm_total) <=0 ) { +perror ("Could not get mbm local bytes" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + +return mbm_total; +} + +pid_t bm_pid, ppid; + +static void ctrlc_handler (int signum,siginfo_t *info,void *ptr) +{ +kill (bm_pid,SIGKILL ); +printf ("Ending\n\n" ); + +exit (EXIT_SUCCESS ); +} + +/* +* print_results_bw: the memory bandwidth results are stored in a file +* @filename: file that stores the results
tools/testing/selftests/resctrl/resctrl_val.c void get_event_and_umask(char *cas_count_cfg, int count, bool op) · line 457
Quoted 10 lines
tools/testing/selftests/resctrl/resctrl_val.c +10
-0+* @bm_pid: child pid that runs benchmark +* @bw_imc: perf imc counter value +* @bw_resc: memory bandwidth value +* +* Return: 0 on success. non-zero on failure. +*/ +static int print_results_bw (char *filename,int bm_pid,float bw_imc, +unsigned long bw_resc) +{ +unsigned long diff =labs (bw_imc - bw_resc);
Since bw_imc is float (why, actually?) this should be fabs(). clang complains about it.
Quoted 57 lines
tools/testing/selftests/resctrl/resctrl_val.c +57
-0+FILE *fp; + +if (strcmp (filename,"stdio" ) ==0 ||strcmp (filename,"stderr" ) ==0 ) { +printf ("Pid: %d \t Mem_BW_iMC: %f \t " , bm_pid, bw_imc); +printf ("Mem_BW_resc: %lu \t Difference: %lu\n" , bw_resc, diff); + }else { + fp =fopen (filename,"a" ); +if (!fp) { +perror ("Cannot open results file" ); + +return errno; + } +if (fprintf (fp,"Pid: %d \t Mem_BW_iMC: %f \t Mem_BW_resc: %lu \t Difference: %lu\n" , + bm_pid, bw_imc, bw_resc, diff) <=0 ) { +fclose (fp); +perror ("Could not log results." ); + +return errno; + } +fclose (fp); + } + +return 0 ; +} + +static int +measure_vals (struct resctrl_val_param *param,unsigned long *bw_resc_start) +{ +unsigned long bw_imc, bw_resc, bw_resc_end; +int ret; + +/* +* Measure memory bandwidth from resctrl and from +* another source which is perf imc value or could +* be something else if perf imc event is not available. +* Compare the two values to validate resctrl value. +* It takes 1sec to measure the data. +*/ + bw_imc =get_mem_bw_imc (param->cpu_no, param->bw_report); +if (bw_imc <=0 ) +return bw_imc; + + bw_resc_end =get_mem_bw_resctrl (); +if (bw_resc_end <=0 ) +return bw_resc_end; + + bw_resc = (bw_resc_end - *bw_resc_start) /MB ; + ret =print_results_bw (param->filename, bm_pid, bw_imc, bw_resc); +if (ret) +return ret; + + *bw_resc_start = bw_resc_end; + +return 0 ; +} + +/*
tools/testing/selftests/resctrl/resctrl_val.c void get_event_and_umask(char *cas_count_cfg, int count, bool op) · line 524
Quoted 10 lines
tools/testing/selftests/resctrl/resctrl_val.c +10
-0+* resctrl_val: execute benchmark and measure memory bandwidth on +* the benchmark +* @benchmark_cmd: benchmark command and its arguments +* @param: parameters passed to resctrl_val() +* +* Return: 0 on success. non-zero on failure. +*/ +int resctrl_val (char **benchmark_cmd,struct resctrl_val_param *param) +{ +int ret =0 , pipefd[2 ], pipe_message =0 ;
If I get this correctly, we just need to pipe for asynchronous signalling. So can't we actually use a signal() call here? At least we should make this pipe_message a single char. This avoids some nasty problems with read() and write() possibly returning early.
Quoted 70 lines
tools/testing/selftests/resctrl/resctrl_val.c +70
-0+char *resctrl_val = param->resctrl_val; +unsigned long bw_resc_start =0 ; +struct sigaction sigact; +union sigval value; +FILE *fp; + +if (strcmp (param->filename,"" ) ==0 ) +sprintf (param->filename,"stdio" ); + +if (strcmp (param->bw_report,"" ) ==0 ) + param->bw_report ="total" ; + +if ((strcmp (resctrl_val,"mba" )) ==0 || + (strcmp (resctrl_val,"mbm" )) ==0 ) { + ret =validate_bw_report_request (param->bw_report); +if (ret) +return ret; + } + + ret =remount_resctrlfs (param->mum_resctrlfs); +if (ret) +return ret; + +/* +* If benchmark wasn't successfully started by child, then child should +* kill parent, so save parent's pid +*/ + ppid =getpid (); + +/* File based synchronization between parent and child */ + fp =fopen ("sig" ,"w" ); +if (!fp) { +perror ("Failed to open sig file" ); + +return -1 ; + } +if (fprintf (fp,"%d\n" ,0 ) <=0 ) { +perror ("Unable to establish sync bw parent & child" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + +if (pipe (pipefd)) { +perror ("Unable to create pipe" ); + +return -1 ; + } + +/* +* Fork to start benchmark, save child's pid so that it can be killed +* when needed +*/ + bm_pid =fork (); +if (bm_pid ==-1 ) { +perror ("Unable to fork" ); + +return -1 ; + } + +if (bm_pid ==0 ) { +/* +* Mask all signals except SIGUSR1, parent uses SIGUSR1 to +* start benchmark +*/ +sigfillset (&sigact.sa_mask); +sigdelset (&sigact.sa_mask,SIGUSR1 ); + + sigact.sa_sigaction = run_benchmark;
tools/testing/selftests/resctrl/resctrl_val.c void get_event_and_umask(char *cas_count_cfg, int count, bool op) · line 604
Quoted 10 lines
tools/testing/selftests/resctrl/resctrl_val.c +10
-0+ sigact.sa_flags =SA_SIGINFO ; + +/* Register for "SIGUSR1" signal from parent */ +if (sigaction (SIGUSR1 , &sigact,NULL )) +PARENT_EXIT ("Can't register child for signal" ); + +/* Tell parent that child is ready */ +close (pipefd[0 ]); + pipe_message =1 ; +write (pipefd[1 ], &pipe_message,sizeof (pipe_message));
gcc -O complains about an unchecked return value of write() here. We could just compare against sizeof(pipe_message) and bail out if it's smaller.
Quoted 41 lines
tools/testing/selftests/resctrl/resctrl_val.c +41
-0+close (pipefd[1 ]); + +/* Suspend child until delivery of "SIGUSR1" from parent */ +sigsuspend (&sigact.sa_mask); + +PARENT_EXIT ("Child is done" ); + } + +printf ("Benchmark PID: %d\n" , bm_pid); + +/* +* Register CTRL-C handler for parent, as it has to kill benchmark +* before exiting +*/ + sigact.sa_sigaction = ctrlc_handler; +sigemptyset (&sigact.sa_mask); + sigact.sa_flags =SA_SIGINFO ; +if (sigaction (SIGINT , &sigact,NULL ) || +sigaction (SIGHUP , &sigact,NULL )) { +perror ("Can't register parent for CTRL-C handler" ); + ret = errno; + goto out; + } + + value.sival_ptr = benchmark_cmd; + +/* Taskset benchmark to specified cpu */ + ret =taskset_benchmark (bm_pid, param->cpu_no); +if (ret) + goto out; + +/* Write benchmark to specified control&monitoring grp in resctrl FS */ + ret =write_bm_pid_to_resctrl (bm_pid, param->ctrlgrp, param->mongrp, + resctrl_val); +if (ret) + goto out; + +if ((strcmp (resctrl_val,"mbm" ) ==0 ) || + (strcmp (resctrl_val,"mba" ) ==0 )) { + ret =initialize_mem_bw_imc (); +if (ret)
tools/testing/selftests/resctrl/resctrl_val.c void get_event_and_umask(char *cas_count_cfg, int count, bool op) · line 655
Quoted 10 lines
tools/testing/selftests/resctrl/resctrl_val.c +10
-0+ goto out; + +initialize_mem_bw_resctrl (param->ctrlgrp, param->mongrp, + param->cpu_no, resctrl_val); + } + +/* Parent waits for child to be ready. */ +close (pipefd[1 ]); +while (pipe_message !=1 ) +read (pipefd[0 ], &pipe_message,sizeof (pipe_message));
Same thing with checking the return value here.
Cheers,
Andre.
Quoted 45 lines
tools/testing/selftests/resctrl/resctrl_val.c +45
-0+close (pipefd[0 ]); + +/* Signal child to start benchmark */ +if (sigqueue (bm_pid,SIGUSR1 , value) ==-1 ) { +perror ("Unable to signal child to start execution" ); + ret = errno; + goto out; + } + +/* Give benchmark enough time to fully run */ +sleep (1 ); + +/* Test runs until the callback setup() tells the test to stop. */ +while (1 ) { +if (strcmp (resctrl_val,"mbm" ) ==0 ) { + ret = param->setup (1 , param); +if (ret) { + ret =0 ; +break ; + } + + ret =measure_vals (param, &bw_resc_start); +if (ret) +break ; + }else if ((strcmp (resctrl_val,"mba" ) ==0 )) { + ret = param->setup (1 , param); +if (ret) { + ret =0 ; +break ; + } + + ret =measure_vals (param, &bw_resc_start); +if (ret) +break ; + }else { +break ; + } + } + +out: +kill (bm_pid,SIGKILL ); +umount_resctrlfs (); + +return ret; +}
PATCH v7 5/13 selftests/resctrl: Add built in benchmark
In reply to Fenghua YuFrom: Sai Praneeth Prakhya <[email protected]>
Built-in benchmark fill_buf generates stressful memory bandwidth and cache traffic.
Later it will be used as a default benchmark by various resctrl tests such as MBA (Memory Bandwidth Allocation) and MBM (Memory Bandwidth Monitoring) tests.
--- tools/testing/selftests/resctrl/fill_buf.c | 204 +++++++++++++++++++++++++++++ 1 file changed, 204 insertions(+) create mode 100644 tools/testing/selftests/resctrl/fill_buf.c
tools/testing/selftests/resctrl/fill_buf.c new file +204 -0
@@ -0,0 +1,204 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* fill_buf benchmark +* +* Copyright (C) 2018 Intel Corporation +* +* Authors: +* Arshiya Hayatkhan Pathan <[email protected]> +* Sai Praneeth Prakhya <[email protected]>, +* Fenghua Yu <[email protected]> +*/ +#include <stdio.h> +#include <unistd.h> +#include <stdlib.h> +#include <sys/types.h> +#include <sys/wait.h> +#include <inttypes.h> +#include <malloc.h> +#include <string.h> + +#include "resctrl.h" + +#define CL_SIZE (64) +#define PAGE_SIZE (4 * 1024) +#define MB (1024 * 1024) + +static unsigned char *startptr; + +static void sb (void ) +{ + asmvolatile ("sfence\n\t" + : : :"memory" ); +} + +static void ctrl_handler (int signo) +{ +free (startptr); +printf ("\nEnding\n" ); +sb (); +exit (EXIT_SUCCESS ); +} + +static void cl_flush (void *p) +{ + asmvolatile ("clflush (%0)\n\t" + : :"r" (p) :"memory" ); +} + +static void mem_flush (void *p,size_t s) +{ +char *cp = (char *)p; +size_t i =0 ; + + s = s /CL_SIZE ;/* mem size in cache llines */ + +for (i =0 ; i < s; i++) +cl_flush (&cp[i *CL_SIZE ]); + +sb (); +} + +static void *malloc_and_init_memory (size_t s) +{ +uint64_t *p64; +size_t s64; + +void *p =memalign (PAGE_SIZE , s); + + p64 = (uint64_t *)p; + s64 = s /sizeof (uint64_t ); + +while (s64 >0 ) { + *p64 = (uint64_t )rand (); + p64 += (CL_SIZE /sizeof (uint64_t )); + s64 -= (CL_SIZE /sizeof (uint64_t )); + } + +return p; +} + +static int fill_one_span_read (unsigned char *start_ptr,unsigned char *end_ptr) +{ +unsigned char sum, *p; + + sum =0 ; + p = start_ptr; +while (p < end_ptr) { + sum += *p; + p += (CL_SIZE /2 ); + } + +return sum; +} + +static +void fill_one_span_write (unsigned char *start_ptr,unsigned char *end_ptr) +{ +unsigned char *p; + + p = start_ptr; +while (p < end_ptr) { + *p ='1' ; + p += (CL_SIZE /2 ); + } +} + +static int fill_cache_read (unsigned char *start_ptr,unsigned char *end_ptr, +char *resctrl_val) +{ +int ret =0 ; +FILE *fp; + +while (1 ) + ret =fill_one_span_read (start_ptr, end_ptr); + +/* Consume read result so that reading memory is not optimized out. */ + fp =fopen ("/dev/null" ,"w" ); +if (!fp) +perror ("Unable to write to /dev/null" ); +fprintf (fp,"Sum: %d " , ret); +fclose (fp); + +return 0 ; +} + +static int fill_cache_write (unsigned char *start_ptr,unsigned char *end_ptr, +char *resctrl_val) +{ +while (1 ) +fill_one_span_write (start_ptr, end_ptr); + +return 0 ; +} + +static int +fill_cache (unsigned long long buf_size,int malloc_and_init,int memflush, +int op,char *resctrl_val) +{ +unsigned char *start_ptr, *end_ptr; +unsigned long long i; +int ret; + +if (malloc_and_init) + start_ptr =malloc_and_init_memory (buf_size); +else + start_ptr =malloc (buf_size); + +if (!start_ptr) +return -1 ; + + startptr = start_ptr; + end_ptr = start_ptr + buf_size; + +/* +* It's better to touch the memory once to avoid any compiler +* optimizations +*/ +if (!malloc_and_init) { +for (i =0 ; i < buf_size; i++) + *start_ptr++ = (unsigned char )rand (); + } + + start_ptr = startptr; + +/* Flush the memory before using to avoid "cache hot pages" effect */ +if (memflush) +mem_flush (start_ptr, buf_size); + +if (op ==0 ) + ret =fill_cache_read (start_ptr, end_ptr, resctrl_val); +else + ret =fill_cache_write (start_ptr, end_ptr, resctrl_val); + +if (ret) { +printf ("\n Errror in fill cache read/write...\n" ); +return -1 ; + } + +free (startptr); + +return 0 ; +} + +int run_fill_buf (unsigned long span,int malloc_and_init_memory, +int memflush,int op,char *resctrl_val) +{ +unsigned long long cache_size = span; +int ret; + +/* set up ctrl-c handler */ +if (signal (SIGINT , ctrl_handler) ==SIG_ERR ) +printf ("Failed to catch SIGINT!\n" ); +if (signal (SIGHUP , ctrl_handler) ==SIG_ERR ) +printf ("Failed to catch SIGHUP!\n" ); + + ret =fill_cache (cache_size, malloc_and_init_memory, memflush, op, + resctrl_val); +if (ret) { +printf ("\n Errror in fill cache\n" ); +return -1 ; + } + +return 0 ; +}
Signature · 1 line
2.7.4
Re: [PATCH v7 05/13] selftests/resctrl: Add built in benchmark
On Sat, 9 Feb 2019 18:50:34 -0800
Fenghua Yu <[email protected]> wrote:
Hi,
Quoted 38 lines
From: Sai Praneeth Prakhya <[email protected]>
Built-in benchmark
fill_bufgenerates stressful memory bandwidth and cache traffic.Later it will be used as a default benchmark by various resctrl tests such as MBA (Memory Bandwidth Allocation) and MBM (Memory Bandwidth Monitoring) tests.
Signed-off-by Sai Praneeth Prakhya <[email protected]>Signed-off-by Arshiya Hayatkhan Pathan <[email protected]>Signed-off-by Fenghua Yu <[email protected]>Signed-off-by Babu Moger <[email protected]>--- tools/testing/selftests/resctrl/fill_buf.c | 204 +++++++++++++++++++++++++++++ 1 file changed, 204 insertions(+) create mode 100644 tools/testing/selftests/resctrl/fill_buf.ctools/testing/selftests/resctrl/fill_buf.c new file +14
-0@@ -0,0 +1,204 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* fill_buf benchmark +* +* Copyright (C) 2018 Intel Corporation +* +* Authors: +* Arshiya Hayatkhan Pathan <[email protected]> +* Sai Praneeth Prakhya <[email protected]>, +* Fenghua Yu <[email protected]> +*/ +#include <stdio.h> +#include <unistd.h> +#include <stdlib.h>
tools/testing/selftests/resctrl/fill_buf.c line 24
Quoted 10 lines
tools/testing/selftests/resctrl/fill_buf.c new file +10
-0+#include <sys/types.h> +#include <sys/wait.h> +#include <inttypes.h> +#include <malloc.h> +#include <string.h> + +#include "resctrl.h" + +#define CL_SIZE (64) +#define PAGE_SIZE (4 * 1024)
This is not really true everywhere. I have definitely seen systems with cache line sizes of 32 and 128 bytes, also arm64 supports 64KB pages. There should be POSIX/glibc functions to read those values at runtime.
tools/testing/selftests/resctrl/fill_buf.c line 32
Quoted 8 lines
tools/testing/selftests/resctrl/fill_buf.c new file +8
-0+#define MB (1024 * 1024) + +static unsigned char *startptr; + +static void sb (void ) +{ + asmvolatile ("sfence\n\t" + : : :"memory" );
Please don't have inline x86 assembly without protecting this code on other architectures. #if defined(__i386) || defined(__x86_64)
does the trick for me.
But aren't there already defined function for those barriers somewhere?
Quoted 4 lines
tools/testing/selftests/resctrl/fill_buf.c new file +4
-0+} + +static void ctrl_handler (int signo) +{
tools/testing/selftests/resctrl/fill_buf.c line 46
Quoted 10 lines
tools/testing/selftests/resctrl/fill_buf.c new file +10
-0+free (startptr); +printf ("\nEnding\n" ); +sb (); +exit (EXIT_SUCCESS ); +} + +static void cl_flush (void *p) +{ + asmvolatile ("clflush (%0)\n\t" + : :"r" (p) :"memory" );
Same comment about the usage of inline assembly here. This breaks compilation on arm/arm64.
In general I am a bit suspicious about using this instruction in a benchmark here, since we have no guarantee that it has an effect on the performance. The CPU could speculatively fetch in data right after this instruction. Shouldn't we iterate multiple times over the benchmark and let the average remove the effect of a hot or cold caches?
Cheers,
Andre.
Quoted 158 lines
tools/testing/selftests/resctrl/fill_buf.c new file +158
-0+} + +static void mem_flush (void *p,size_t s) +{ +char *cp = (char *)p; +size_t i =0 ; + + s = s /CL_SIZE ;/* mem size in cache llines */ + +for (i =0 ; i < s; i++) +cl_flush (&cp[i *CL_SIZE ]); + +sb (); +} + +static void *malloc_and_init_memory (size_t s) +{ +uint64_t *p64; +size_t s64; + +void *p =memalign (PAGE_SIZE , s); + + p64 = (uint64_t *)p; + s64 = s /sizeof (uint64_t ); + +while (s64 >0 ) { + *p64 = (uint64_t )rand (); + p64 += (CL_SIZE /sizeof (uint64_t )); + s64 -= (CL_SIZE /sizeof (uint64_t )); + } + +return p; +} + +static int fill_one_span_read (unsigned char *start_ptr,unsigned char *end_ptr) +{ +unsigned char sum, *p; + + sum =0 ; + p = start_ptr; +while (p < end_ptr) { + sum += *p; + p += (CL_SIZE /2 ); + } + +return sum; +} + +static +void fill_one_span_write (unsigned char *start_ptr,unsigned char *end_ptr) +{ +unsigned char *p; + + p = start_ptr; +while (p < end_ptr) { + *p ='1' ; + p += (CL_SIZE /2 ); + } +} + +static int fill_cache_read (unsigned char *start_ptr,unsigned char *end_ptr, +char *resctrl_val) +{ +int ret =0 ; +FILE *fp; + +while (1 ) + ret =fill_one_span_read (start_ptr, end_ptr); + +/* Consume read result so that reading memory is not optimized out. */ + fp =fopen ("/dev/null" ,"w" ); +if (!fp) +perror ("Unable to write to /dev/null" ); +fprintf (fp,"Sum: %d " , ret); +fclose (fp); + +return 0 ; +} + +static int fill_cache_write (unsigned char *start_ptr,unsigned char *end_ptr, +char *resctrl_val) +{ +while (1 ) +fill_one_span_write (start_ptr, end_ptr); + +return 0 ; +} + +static int +fill_cache (unsigned long long buf_size,int malloc_and_init,int memflush, +int op,char *resctrl_val) +{ +unsigned char *start_ptr, *end_ptr; +unsigned long long i; +int ret; + +if (malloc_and_init) + start_ptr =malloc_and_init_memory (buf_size); +else + start_ptr =malloc (buf_size); + +if (!start_ptr) +return -1 ; + + startptr = start_ptr; + end_ptr = start_ptr + buf_size; + +/* +* It's better to touch the memory once to avoid any compiler +* optimizations +*/ +if (!malloc_and_init) { +for (i =0 ; i < buf_size; i++) + *start_ptr++ = (unsigned char )rand (); + } + + start_ptr = startptr; + +/* Flush the memory before using to avoid "cache hot pages" effect */ +if (memflush) +mem_flush (start_ptr, buf_size); + +if (op ==0 ) + ret =fill_cache_read (start_ptr, end_ptr, resctrl_val); +else + ret =fill_cache_write (start_ptr, end_ptr, resctrl_val); + +if (ret) { +printf ("\n Errror in fill cache read/write...\n" ); +return -1 ; + } + +free (startptr); + +return 0 ; +} + +int run_fill_buf (unsigned long span,int malloc_and_init_memory, +int memflush,int op,char *resctrl_val) +{ +unsigned long long cache_size = span; +int ret; + +/* set up ctrl-c handler */ +if (signal (SIGINT , ctrl_handler) ==SIG_ERR ) +printf ("Failed to catch SIGINT!\n" ); +if (signal (SIGHUP , ctrl_handler) ==SIG_ERR ) +printf ("Failed to catch SIGHUP!\n" ); + + ret =fill_cache (cache_size, malloc_and_init_memory, memflush, op, + resctrl_val); +if (ret) { +printf ("\n Errror in fill cache\n" ); +return -1 ; + } + +return 0 ; +}
PATCH v7 6/13 selftests/resctrl: Add MBM test
In reply to Fenghua YuFrom: Arshiya Hayatkhan Pathan <[email protected]>
MBM (Memory Bandwidth Monitoring) test is the first implemented selftest. It starts a stressful memory bandwidth benchmark and assigns the bandwidth pid in a resctrl monitoring group. Read and compare perf IMC counter and MBM total bytes for the benchmark. The numbers should be close enough to pass the test.
Default benchmark is built-in fill_buf. But users can specify their own benchmark by option "-b".
We can add memory bandwidth monitoring for multiple processes in the future.
--- tools/testing/selftests/resctrl/Makefile | 8 +- tools/testing/selftests/resctrl/mbm_test.c | 149 ++++++++++++++++++++++++ tools/testing/selftests/resctrl/resctrl.h | 3 + tools/testing/selftests/resctrl/resctrl_tests.c | 114 ++++++++++++++++++ tools/testing/selftests/resctrl/resctrl_val.c | 2 + 5 files changed, 275 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/resctrl/mbm_test.c create mode 100644 tools/testing/selftests/resctrl/resctrl_tests.c
tools/testing/selftests/resctrl/Makefile +7 -1
@@ -1,10 +1,16 @@ CC = gcc CFLAGS = -g -Wall +all: resctrl_tests + *.o: *.c $(CC) $(CFLAGS) -c *.c +resctrl_tests: *.o + $(CC) $(CFLAGS) -o resctrl_tests resctrl_tests.o resctrlfs.o \ + resctrl_val.o fill_buf.o mbm_test.o + .PHONY: clean clean: - $(RM) *.o *~ + $(RM) *.o *~ resctrl_tests
tools/testing/selftests/resctrl/mbm_test.c new file +149 -0
149 lines not on this page · raw message
tools/testing/selftests/resctrl/resctrl.h +3 -0
@@ -71,5 +71,8 @@ int perf_event_open(struct perf_event_attr *hw_event, pid_t pid, int cpu,int run_fill_buf (unsigned long span,int malloc_and_init_memory,int memflush,int op,char *resctrl_va);int resctrl_val (char **benchmark_cmd,struct resctrl_val_param *param); +int mbm_bw_change (int span,int cpu_no,char *bw_report,char **benchmark_cmd); +void tests_cleanup (void ); +void mbm_test_cleanup (void );#endif /* RESCTRL_H */
tools/testing/selftests/resctrl/resctrl_tests.c new file +114 -0
114 lines not on this page · raw message
tools/testing/selftests/resctrl/resctrl_val.c +2 -0
@@ -437,6 +437,8 @@ pid_t bm_pid, ppid;static void ctrlc_handler (int signum,siginfo_t *info,void *ptr) {kill (bm_pid,SIGKILL ); +umount_resctrlfs (); +tests_cleanup ();printf ("Ending\n\n" );exit (EXIT_SUCCESS );
Signature · 1 line
2.7.4
Re: [PATCH v7 06/13] selftests/resctrl: Add MBM test
On Sat, 9 Feb 2019 18:50:35 -0800
Fenghua Yu <[email protected]> wrote:
Hi,
Quoted 34 lines
From: Arshiya Hayatkhan Pathan <[email protected]>
MBM (Memory Bandwidth Monitoring) test is the first implemented selftest. It starts a stressful memory bandwidth benchmark and assigns the bandwidth pid in a resctrl monitoring group. Read and compare perf IMC counter and MBM total bytes for the benchmark. The numbers should be close enough to pass the test.
Default benchmark is built-in
fill_buf. But users can specify their own benchmark by option "-b".We can add memory bandwidth monitoring for multiple processes in the future.
Signed-off-by Arshiya Hayatkhan Pathan <[email protected]>Signed-off-by Sai Praneeth Prakhya <[email protected]>,Signed-off-by Fenghua Yu <[email protected]>Signed-off-by Babu Moger <[email protected]>--- tools/testing/selftests/resctrl/Makefile | 8 +- tools/testing/selftests/resctrl/mbm_test.c | 149 ++++++++++++++++++++++++ tools/testing/selftests/resctrl/resctrl.h | 3 + tools/testing/selftests/resctrl/resctrl_tests.c | 114 ++++++++++++++++++ tools/testing/selftests/resctrl/resctrl_val.c | 2 + 5 files changed, 275 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/resctrl/mbm_test.c create mode 100644 tools/testing/selftests/resctrl/resctrl_tests.ctools/testing/selftests/resctrl/Makefile +0
-0@@ -1,10 +1,16 @@ CC = gcc
tools/testing/selftests/resctrl/Makefile line 11
Quoted 10 lines
tools/testing/selftests/resctrl/Makefile +5
-0CFLAGS = -g -Wall +all: resctrl_tests + *.o: *.c $(CC) $(CFLAGS) -c *.c +resctrl_tests: *.o + $(CC) $(CFLAGS) -o resctrl_tests resctrl_tests.o resctrlfs.o \ + resctrl_val.o fill_buf.o mbm_test.o
This is a somewhat non-standard way of specifying things in a Makefile. What about:
OBJS = resctrl_tests.o resctrlfs.o resctrl_val.o fill_buf.o mbm_test.oresctrl_tests: $(OBJS)
$(CC) $(LDFLAGS) -o $@ $^
Quoted 17 lines
tools/testing/selftests/resctrl/Makefile +2
-1+ .PHONY: clean clean: - $(RM) *.o *~ + $(RM) *.o *~ resctrl_teststools/testing/selftests/resctrl/mbm_test.c new file +5
-0@@ -0,0 +1,149 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Memory Bandwidth Monitoring (MBM) test +* +* Copyright (C) 2018 Intel Corporation
tools/testing/selftests/resctrl/mbm_test.c line 15
Quoted 10 lines
tools/testing/selftests/resctrl/mbm_test.c new file +10
-0+* +* Authors: +* Arshiya Hayatkhan Pathan <[email protected]> +* Sai Praneeth Prakhya <[email protected]>, +* Fenghua Yu <[email protected]> +*/ +#include "resctrl.h" + +#define RESULT_FILE_NAME "result_mbm" +#define MAX_DIFF 300
What is that value? 300 MB/s? Is that some heuristic? Shouldn't that be relative, like in the other benchmarks?
tools/testing/selftests/resctrl/mbm_test.c line 21
Quoted 6 lines
tools/testing/selftests/resctrl/mbm_test.c new file +6
-0+#define NUM_OF_RUNS 5 + +static void +show_bw_info (unsigned long *bw_imc,unsigned long *bw_resc,int span) +{ +unsigned long avg_bw_imc =0 , avg_bw_resc =0 ;
Why do you initialise those if they are overwritten below?
tools/testing/selftests/resctrl/mbm_test.c line 23
Quoted 2 lines
tools/testing/selftests/resctrl/mbm_test.c new file +2
-0+unsigned long sum_bw_imc =0 , sum_bw_resc =0 ; +long avg_diff =0 ;
Same here.
Quoted 4 lines
tools/testing/selftests/resctrl/mbm_test.c new file +4
-0+int runs; + +/* +* Discard the first value which is inaccurate due to monitoring setup
tools/testing/selftests/resctrl/mbm_test.c line 37
Quoted 10 lines
tools/testing/selftests/resctrl/mbm_test.c new file +10
-0+* transition phase. +*/ +for (runs =1 ; runs <NUM_OF_RUNS ; runs++) { + sum_bw_imc += bw_imc[runs]; + sum_bw_resc += bw_resc[runs]; + } + + avg_bw_imc = sum_bw_imc /4 ; + avg_bw_resc = sum_bw_resc /4 ; + avg_diff = avg_bw_resc - avg_bw_imc;
You could make avg_diff unsigned and use the labs() call already here.
Cheers,
Andre.
Quoted 258 lines
tools/testing/selftests/resctrl/mbm_test.c new file +112
-0+ +printf ("\nSpan (MB): %d \t" , span); +printf ("avg_bw_imc: %lu\t" , avg_bw_imc); +printf ("avg_bw_resc: %lu\t" , avg_bw_resc); +printf ("avg_diff: %lu\t" ,labs (avg_diff)); + +if (labs (avg_diff) >MAX_DIFF ) +printf (" failed\n" ); +else +printf (" passed\n" ); +} + +static int check_results (int span) +{ +unsigned long bw_imc[NUM_OF_RUNS ], bw_resc[NUM_OF_RUNS ]; +char temp[1024 ], *token_array[8 ]; +char output[] =RESULT_FILE_NAME ; +int runs; +FILE *fp; + +printf ("\nChecking for pass/fail\n" ); + + fp =fopen (output,"r" ); +if (!fp) { +perror ("Error in opening file\n" ); + +return errno; + } + + runs =0 ; +while (fgets (temp,1024 , fp)) { +char *token =strtok (temp,":\t" ); +int i =0 ; + +while (token) { + token_array[i++] = token; + token =strtok (NULL ,":\t" ); + } + + bw_resc[runs] =strtoul (token_array[5 ],NULL ,0 ); + bw_imc[runs] =strtoul (token_array[3 ],NULL ,0 ); + runs++; + } + +show_bw_info (bw_imc, bw_resc, span); + +fclose (fp); + +return 0 ; +} + +static int mbm_setup (int num, ...) +{ +struct resctrl_val_param *p; +static int num_of_runs; +va_list param; +int ret =0 ; + +/* Run NUM_OF_RUNS times */ +if (num_of_runs++ >=NUM_OF_RUNS ) +return -1 ; + +va_start (param, num); + p = va_arg(param, struct resctrl_val_param *); +va_end (param); + +/* Set up shemata with 100% allocation on the first run. */ +if (num_of_runs ==0 ) + ret =write_schemata (p->ctrlgrp,"100" , p->cpu_no, + p->resctrl_val); + +return ret; +} + +void mbm_test_cleanup (void ) +{ +remove (RESULT_FILE_NAME ); +} + +int mbm_bw_change (int span,int cpu_no,char *bw_report,char **benchmark_cmd) +{ +struct resctrl_val_param param = { + .resctrl_val ="mbm" , + .ctrlgrp ="c1" , + .mongrp ="m1" , + .span = span, + .cpu_no = cpu_no, + .mum_resctrlfs =1 , + .filename =RESULT_FILE_NAME , + .bw_report = bw_report, + .setup = mbm_setup + }; +int ret; + +remove (RESULT_FILE_NAME ); + + ret =validate_resctrl_feature_request ("mbm" ); +if (ret) +return ret; + + ret =resctrl_val (benchmark_cmd, ¶m); +if (ret) +return ret; + + ret =check_results (span); +if (ret) +return ret; + +mbm_test_cleanup (); + +return 0 ; +}tools/testing/selftests/resctrl/resctrl.h +3
-0@@ -71,5 +71,8 @@ int perf_event_open(struct perf_event_attr *hw_event, pid_t pid, int cpu,int run_fill_buf (unsigned long span,int malloc_and_init_memory,int memflush,int op,char *resctrl_va);int resctrl_val (char **benchmark_cmd,struct resctrl_val_param *param); +int mbm_bw_change (int span,int cpu_no,char *bw_report,char **benchmark_cmd); +void tests_cleanup (void ); +void mbm_test_cleanup (void );#endif /* RESCTRL_H */ tools/testing/selftests/resctrl/resctrl_tests.c new file +114
-0@@ -0,0 +1,114 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Resctrl tests +* +* Copyright (C) 2018 Intel Corporation +* +* Authors: +* Arshiya Hayatkhan Pathan <[email protected]> +* Sai Praneeth Prakhya <[email protected]>, +* Fenghua Yu <[email protected]> +*/ +#include "resctrl.h" + +#define BENCHMARK_ARGS 64 +#define BENCHMARK_ARG_SIZE 64 + +static void cmd_help (void ) +{ +printf ("usage: resctrl_tests [-h] [-b \"benchmark_cmd [options]\"] [-t test list]\n" ); +printf ("\t-b benchmark_cmd [options]: run specified benchmark\n" ); +printf ("\t default benchmark is builtin fill_buf\n" ); +printf ("\t-t test list: run tests specified in the test list, " ); +printf ("e.g. -t mbm,mba\n" ); +printf ("\t-h: help\n" ); +} + +void tests_cleanup (void ) +{ +mbm_test_cleanup (); +} + +int main (int argc,char **argv) +{ +char benchmark_cmd_area[BENCHMARK_ARGS ][BENCHMARK_ARG_SIZE ]; +int res, c, cpu_no =1 , span =250 , argc_new = argc, i; +int ben_ind; +bool has_ben = false, mbm_test = true; +char *benchmark_cmd[BENCHMARK_ARGS ]; +char bw_report[64 ], bm_type[64 ]; + +for (i =0 ; i < argc; i++) { +if (strcmp (argv[i],"-b" ) ==0 ) { + ben_ind = i +1 ; + argc_new = ben_ind -1 ; + has_ben =1 ; +break ; + } + } + +while ((c =getopt (argc_new, argv,"ht:b:" )) !=-1 ) { +char *token; + +switch (c) { +case 't' : + token =strtok (optarg,"," ); + + mbm_test = false; +while (token) { +if (!strcmp (token,"mbm" )) { + mbm_test = true; + }else { +printf ("invalid argument\n" ); + +return -1 ; + } + token =strtok (NULL ,":\t" ); + } +break ; +case 'h' : +cmd_help (); + +return 0 ; +default : +printf ("invalid argument\n" ); + +return -1 ; + } + } + +/* +* We need root privileges to run because +* 1. We write to resctrl FS +* 2. We execute perf commands +*/ +if (geteuid () !=0 ) { +perror ("Please run this program as root\n" ); + +return errno; + } + +if (!has_ben) { +/* If no benchmark is given by "-b" argument, use fill_buf. */ +for (i =0 ; i <5 ; i++) + benchmark_cmd[i] = benchmark_cmd_area[i]; +strcpy (benchmark_cmd[0 ],"fill_buf" ); +sprintf (benchmark_cmd[1 ],"%d" , span); +strcpy (benchmark_cmd[2 ],"1" ); +strcpy (benchmark_cmd[3 ],"1" ); +strcpy (benchmark_cmd[4 ],"0" ); + benchmark_cmd[5 ] =NULL ; + } + +sprintf (bw_report,"reads" ); +sprintf (bm_type,"fill_buf" ); + +if (mbm_test) { +printf ("\nMBM BW Change Starting..\n" ); + res =mbm_bw_change (span, cpu_no, bw_report, benchmark_cmd); +if (res) +printf ("Error in running tests for mbm bw change!\n" ); + } + +return 0 ; +}tools/testing/selftests/resctrl/resctrl_val.c +2
-0@@ -437,6 +437,8 @@ pid_t bm_pid, ppid;static void ctrlc_handler (int signum,siginfo_t *info,void *ptr) {kill (bm_pid,SIGKILL ); +umount_resctrlfs (); +tests_cleanup ();printf ("Ending\n\n" );exit (EXIT_SUCCESS );
PATCH v7 7/13 selftests/resctrl: Add MBA test
In reply to Fenghua YuFrom: Arshiya Hayatkhan Pathan <[email protected]>
MBA (Memory Bandwidth Allocation) test starts a stressful memory bandwidth benchmark and allocates memory bandwidth from 100% down to 10% for the benchmark process. For each allocation, compare perf IMC counter and mbm total bytes from resctrl. The difference between the two values should be within a threshold to pass the test.
Default benchmark is built-in fill_buf. But users can specify their own benchmark by option "-b".
We can add memory bandwidth allocation for multiple processes in the future.
--- tools/testing/selftests/resctrl/Makefile | 2 +- tools/testing/selftests/resctrl/mba_test.c | 177 ++++++++++++++++++++++++ tools/testing/selftests/resctrl/resctrl.h | 2 + tools/testing/selftests/resctrl/resctrl_tests.c | 15 +- 4 files changed, 194 insertions(+), 2 deletions(-) create mode 100644 tools/testing/selftests/resctrl/mba_test.c
tools/testing/selftests/resctrl/Makefile +1 -1
@@ -8,7 +8,7 @@ all: resctrl_tests resctrl_tests: *.o $(CC) $(CFLAGS) -o resctrl_tests resctrl_tests.o resctrlfs.o \ - resctrl_val.o fill_buf.o mbm_test.o + resctrl_val.o fill_buf.o mbm_test.o mba_test.o .PHONY: clean
tools/testing/selftests/resctrl/mba_test.c new file +177 -0
177 lines not on this page · raw message
tools/testing/selftests/resctrl/resctrl.h +2 -0
@@ -74,5 +74,7 @@ int resctrl_val(char **benchmark_cmd, struct resctrl_val_param *param);int mbm_bw_change (int span,int cpu_no,char *bw_report,char **benchmark_cmd);void tests_cleanup (void );void mbm_test_cleanup (void ); +int mba_schemata_change (int cpu_no,char *bw_report,char **benchmark_cmd); +void mba_test_cleanup (void );#endif /* RESCTRL_H */
tools/testing/selftests/resctrl/resctrl_tests.c +14 -1
@@ -27,6 +27,7 @@ static void cmd_help(void)void tests_cleanup (void ) {mbm_test_cleanup (); +mba_test_cleanup (); }int main (int argc,char **argv)
@@ -34,7 +35,7 @@ int main(int argc, char **argv)char benchmark_cmd_area[BENCHMARK_ARGS ][BENCHMARK_ARG_SIZE ];int res, c, cpu_no =1 , span =250 , argc_new = argc, i;int ben_ind; -bool has_ben = false, mbm_test = true; +bool has_ben = false, mbm_test = true, mba_test = true;char *benchmark_cmd[BENCHMARK_ARGS ];char bw_report[64 ], bm_type[64 ];
@@ -55,9 +56,12 @@ int main(int argc, char **argv) token =strtok (optarg,"," ); mbm_test = false; + mba_test = false;while (token) {if (!strcmp (token,"mbm" )) { mbm_test = true; + }else if (!strcmp (token,"mba" )) { + mba_test = true; }else {printf ("invalid argument\n" );
@@ -110,5 +114,14 @@ int main(int argc, char **argv)printf ("Error in running tests for mbm bw change!\n" ); } +if (mba_test) { +printf ("\nMBA Schemata Change Starting..\n" ); +if (!has_ben) +sprintf (benchmark_cmd[1 ],"%d" , span); + res =mba_schemata_change (cpu_no, bw_report, benchmark_cmd); +if (res) +printf ("Error in tests for mba-change-schemata!\n" ); + } +return 0 ; }
Signature · 1 line
2.7.4
Re: [PATCH v7 07/13] selftests/resctrl: Add MBA test
On Sat, 9 Feb 2019 18:50:36 -0800
Fenghua Yu <[email protected]> wrote:
Hi,
Quoted 51 lines
From: Arshiya Hayatkhan Pathan <[email protected]>
MBA (Memory Bandwidth Allocation) test starts a stressful memory bandwidth benchmark and allocates memory bandwidth from 100% down to 10% for the benchmark process. For each allocation, compare perf IMC counter and mbm total bytes from resctrl. The difference between the two values should be within a threshold to pass the test.
Default benchmark is built-in
fill_buf. But users can specify their own benchmark by option "-b".We can add memory bandwidth allocation for multiple processes in the future.
Signed-off-by Arshiya Hayatkhan Pathan <[email protected]>Signed-off-by Sai Praneeth Prakhya <[email protected]>,Signed-off-by Fenghua Yu <[email protected]>Signed-off-by Babu Moger <[email protected]>--- tools/testing/selftests/resctrl/Makefile | 2 +- tools/testing/selftests/resctrl/mba_test.c | 177 ++++++++++++++++++++++++ tools/testing/selftests/resctrl/resctrl.h | 2 + tools/testing/selftests/resctrl/resctrl_tests.c | 15 +- 4 files changed, 194 insertions(+), 2 deletions(-) create mode 100644 tools/testing/selftests/resctrl/mba_test.ctools/testing/selftests/resctrl/Makefile +1
-1@@ -8,7 +8,7 @@ all: resctrl_tests resctrl_tests: *.o $(CC) $(CFLAGS) -o resctrl_tests resctrl_tests.o resctrlfs.o \ - resctrl_val.o fill_buf.o mbm_test.o + resctrl_val.o fill_buf.o mbm_test.o mba_test.o .PHONY: cleantools/testing/selftests/resctrl/mba_test.c new file +6
-0@@ -0,0 +1,177 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Memory Bandwidth Allocation (MBA) test +* +* Copyright (C) 2018 Intel Corporation +*
tools/testing/selftests/resctrl/mba_test.c line 16
Quoted 10 lines
tools/testing/selftests/resctrl/mba_test.c new file +10
-0+* Authors: +* Arshiya Hayatkhan Pathan <[email protected]> +* Sai Praneeth Prakhya <[email protected]>, +* Fenghua Yu <[email protected]> +*/ +#include "resctrl.h" + +#define RESULT_FILE_NAME "result_mba" +#define NUM_OF_RUNS 5 +#define MAX_DIFF 300
What is the unit here? MB/s?
Quoted 37 lines
tools/testing/selftests/resctrl/mba_test.c new file +37
-0+#define ALLOCATION_MAX 100 +#define ALLOCATION_MIN 10 +#define ALLOCATION_STEP 10 + +/* +* Change schemata percentage from 100 to 10%. Write schemata to specified +* con_mon grp, mon_grp in resctrl FS. +* For each allocation, run 5 times in order to get average values. +*/ +static int mba_setup (int num, ...) +{ +static int runs_per_allocation, allocation =100 ; +struct resctrl_val_param *p; +char allocation_str[64 ]; +va_list param; + +va_start (param, num); + p = va_arg(param, struct resctrl_val_param *); +va_end (param); + +if (runs_per_allocation >=NUM_OF_RUNS ) + runs_per_allocation =0 ; + +/* Only set up schemata once every NUM_OF_RUNS of allocations */ +if (runs_per_allocation++ !=0 ) +return 0 ; + +if (allocation <ALLOCATION_MIN || allocation >ALLOCATION_MAX ) +return -1 ; + +sprintf (allocation_str,"%d" , allocation); + +write_schemata (p->ctrlgrp, allocation_str, p->cpu_no, p->resctrl_val); +printf ("changed schemata to : %d\n" , allocation); + allocation -=ALLOCATION_STEP ; + +return 0 ;
tools/testing/selftests/resctrl/mba_test.c line 63
Quoted 10 lines
tools/testing/selftests/resctrl/mba_test.c new file +10
-0+} + +static void show_mba_info (unsigned long *bw_imc,unsigned long *bw_resc) +{ +int allocation, failed =0 , runs; + +/* Memory bandwidth from 100% down to 10% */ +for (allocation =0 ; allocation <ALLOCATION_MAX /ALLOCATION_STEP ; + allocation++) { +unsigned long avg_bw_imc =0 , avg_bw_resc =0 ;
No need to initialise those.
tools/testing/selftests/resctrl/mba_test.c line 65
Quoted 2 lines
tools/testing/selftests/resctrl/mba_test.c new file +2
-0+unsigned long sum_bw_imc =0 , sum_bw_resc =0 ; +long avg_diff =0 ;
... and this one.
Quoted 4 lines
tools/testing/selftests/resctrl/mba_test.c new file +4
-0+ +/* +* The first run is discarded due to inaccurate value from +* phase transition.
tools/testing/selftests/resctrl/mba_test.c line 79
Quoted 10 lines
tools/testing/selftests/resctrl/mba_test.c new file +10
-0+*/ +for (runs =NUM_OF_RUNS * allocation +1 ; + runs <NUM_OF_RUNS * allocation +NUM_OF_RUNS ; runs++) { + sum_bw_imc += bw_imc[runs]; + sum_bw_resc += bw_resc[runs]; + } + + avg_bw_imc = sum_bw_imc / (NUM_OF_RUNS -1 ); + avg_bw_resc = sum_bw_resc / (NUM_OF_RUNS -1 ); + avg_diff = avg_bw_resc - avg_bw_imc;
Same comment as in the last patch: use labs() here already.
Cheers,
Andre.
Quoted 15 lines
tools/testing/selftests/resctrl/mba_test.c new file +15
-0+ +printf ("\nschemata percentage: %d \t" , +ALLOCATION_MAX -ALLOCATION_STEP * allocation); +printf ("avg_bw_imc: %lu\t" , avg_bw_imc); +printf ("avg_bw_resc: %lu\t" , avg_bw_resc); +printf ("avg_diff: %lu\t" ,labs (avg_diff)); +if (labs (avg_diff) >MAX_DIFF ) { +printf ("failed\n" ); + failed =1 ; + }else { +printf ("passed\n" ); + } + } + +if (failed) {
tools/testing/selftests/resctrl/mba_test.c line 104
Quoted 10 lines
tools/testing/selftests/resctrl/mba_test.c new file +10
-0+printf ("\nTest for schemata change using MBA failed" ); +printf ("as atleast one test failed!\n" ); + }else { +printf ("\nTests for changing schemata using MBA passed!\n\n" ); + } +} + +static int check_results (void ) +{ +char *token_array[8 ], output[] =RESULT_FILE_NAME , temp[512 ];
The size of temp[] does not match below the usage ...
Quoted 4 lines
tools/testing/selftests/resctrl/mba_test.c new file +4
-0+unsigned long bw_imc[1024 ], bw_resc[1024 ]; +int runs; +FILE *fp; +
tools/testing/selftests/resctrl/mba_test.c line 118
Quoted 10 lines
tools/testing/selftests/resctrl/mba_test.c new file +10
-0+printf ("\nchecking for pass/fail\n" ); + fp =fopen (output,"r" ); +if (!fp) { +perror ("Error in opening file\n" ); + +return errno; + } + + runs =0 ; +while (fgets (temp,1024 , fp)) {
... here. You could just use sizeof(temp) here.
Quoted 120 lines
tools/testing/selftests/resctrl/mba_test.c new file +59
-0+char *token =strtok (temp,":\t" ); +int fields =0 ; + +while (token) { + token_array[fields++] = token; + token =strtok (NULL ,":\t" ); + } + +/* Field 3 is perf imc value */ + bw_imc[runs] =strtoul (token_array[3 ],NULL ,0 ); +/* Field 5 is resctrl value */ + bw_resc[runs] =strtoul (token_array[5 ],NULL ,0 ); + runs++; + } + +fclose (fp); + +show_mba_info (bw_imc, bw_resc); + +return 0 ; +} + +void mba_test_cleanup (void ) +{ +remove (RESULT_FILE_NAME ); +} + +int mba_schemata_change (int cpu_no,char *bw_report,char **benchmark_cmd) +{ +struct resctrl_val_param param = { + .resctrl_val ="mba" , + .ctrlgrp ="c1" , + .mongrp ="m1" , + .cpu_no = cpu_no, + .mum_resctrlfs =1 , + .filename =RESULT_FILE_NAME , + .bw_report = bw_report, + .setup = mba_setup + }; +int ret; + +remove (RESULT_FILE_NAME ); + + ret =validate_resctrl_feature_request ("mba" ); +if (ret) +return ret; + + ret =resctrl_val (benchmark_cmd, ¶m); +if (ret) +return ret; + + ret =check_results (); +if (ret) +return ret; + +mba_test_cleanup (); + +return 0 ; +}tools/testing/selftests/resctrl/resctrl.h +2
-0@@ -74,5 +74,7 @@ int resctrl_val(char **benchmark_cmd, struct resctrl_val_param *param);int mbm_bw_change (int span,int cpu_no,char *bw_report,char **benchmark_cmd);void tests_cleanup (void );void mbm_test_cleanup (void ); +int mba_schemata_change (int cpu_no,char *bw_report,char **benchmark_cmd); +void mba_test_cleanup (void );#endif /* RESCTRL_H */ tools/testing/selftests/resctrl/resctrl_tests.c +14
-1@@ -27,6 +27,7 @@ static void cmd_help(void)void tests_cleanup (void ) {mbm_test_cleanup (); +mba_test_cleanup (); }int main (int argc,char **argv)@@ -34,7 +35,7 @@ int main(int argc, char **argv)char benchmark_cmd_area[BENCHMARK_ARGS ][BENCHMARK_ARG_SIZE ];int res, c, cpu_no =1 , span =250 , argc_new = argc, i;int ben_ind; -bool has_ben = false, mbm_test = true; +bool has_ben = false, mbm_test = true, mba_test = true;char *benchmark_cmd[BENCHMARK_ARGS ];char bw_report[64 ], bm_type[64 ];@@ -55,9 +56,12 @@ int main(int argc, char **argv) token =strtok (optarg,"," ); mbm_test = false; + mba_test = false;while (token) {if (!strcmp (token,"mbm" )) { mbm_test = true; + }else if (!strcmp (token,"mba" )) { + mba_test = true; }else {printf ("invalid argument\n" );@@ -110,5 +114,14 @@ int main(int argc, char **argv)printf ("Error in running tests for mbm bw change!\n" ); } +if (mba_test) { +printf ("\nMBA Schemata Change Starting..\n" ); +if (!has_ben) +sprintf (benchmark_cmd[1 ],"%d" , span); + res =mba_schemata_change (cpu_no, bw_report, benchmark_cmd); +if (res) +printf ("Error in tests for mba-change-schemata!\n" ); + } +return 0 ; }
PATCH v7 8/13 selftests/resctrl: Add Cache QoS Monitoring (CQM) selftest
In reply to Fenghua YuFrom: Arshiya Hayatkhan Pathan <[email protected]>
Cache QoS Monitoring (CQM) selftest starts stressful cache benchmark with specified size of memory to access the cache. Last Level cache occupancy reported by CQM should be close to the size of the memory.
--- tools/testing/selftests/resctrl/Makefile | 2 +- tools/testing/selftests/resctrl/cache.c | 101 +++++++++++ tools/testing/selftests/resctrl/cqm_test.c | 173 ++++++++++++++++++ tools/testing/selftests/resctrl/mba_test.c | 1 + tools/testing/selftests/resctrl/mbm_test.c | 1 + tools/testing/selftests/resctrl/resctrl.h | 23 ++- tools/testing/selftests/resctrl/resctrl_tests.c | 56 ++++-- tools/testing/selftests/resctrl/resctrl_val.c | 95 ++++++---- tools/testing/selftests/resctrl/resctrlfs.c | 227 ++++++++++++++++++++++-- 9 files changed, 614 insertions(+), 65 deletions(-) create mode 100644 tools/testing/selftests/resctrl/cache.c create mode 100644 tools/testing/selftests/resctrl/cqm_test.c
tools/testing/selftests/resctrl/Makefile +1 -1
@@ -8,7 +8,7 @@ all: resctrl_tests resctrl_tests: *.o $(CC) $(CFLAGS) -o resctrl_tests resctrl_tests.o resctrlfs.o \ - resctrl_val.o fill_buf.o mbm_test.o mba_test.o + resctrl_val.o fill_buf.o mbm_test.o mba_test.o cache.o cqm_test.o .PHONY: clean
tools/testing/selftests/resctrl/cache.c new file +101 -0
101 lines not on this page · raw message
tools/testing/selftests/resctrl/cqm_test.c new file +173 -0
173 lines not on this page · raw message
tools/testing/selftests/resctrl/mba_test.c +1 -0
@@ -57,6 +57,7 @@ static void show_mba_info(unsigned long *bw_imc, unsigned long *bw_resc) {int allocation, failed =0 , runs; +printf ("\nResults are displayed in (MB)\n" );/* Memory bandwidth from 100% down to 10% */ for (allocation =0 ; allocation <ALLOCATION_MAX /ALLOCATION_STEP ; allocation++) {
tools/testing/selftests/resctrl/mbm_test.c +1 -0
@@ -36,6 +36,7 @@ show_bw_info(unsigned long *bw_imc, unsigned long *bw_resc, int span) avg_bw_resc = sum_bw_resc /4 ; avg_diff = avg_bw_resc - avg_bw_imc; +printf ("\nResults are displayed in (MB)\n" );printf ("\nSpan (MB): %d \t" , span);printf ("avg_bw_imc: %lu\t" , avg_bw_imc);printf ("avg_bw_resc: %lu\t" , avg_bw_resc);
tools/testing/selftests/resctrl/resctrl.h +22 -1
@@ -16,11 +16,18 @@#include <sys/ioctl.h> #include <sys/mount.h> #include <sys/types.h> +#include <sys/select.h> #include <asm/unistd.h> #include <linux/perf_event.h> +#include <sys/time.h> +#include <math.h> +#include <sys/wait.h> +#include <sys/eventfd.h> +#define MB (1024 * 1024)#define RESCTRL_PATH "/sys/fs/resctrl"#define PHYS_ID_PATH "/sys/devices/system/cpu/cpu" +#define CBM_MASK_PATH "/sys/fs/resctrl/info"#define PARENT_EXIT (err_msg) \ do { \
@@ -46,14 +53,20 @@ struct resctrl_val_param {char ctrlgrp[64 ];char mongrp[64 ];int cpu_no; -int span; +unsigned long long span;int mum_resctrlfs;char filename[64 ];char *bw_report; +unsigned long mask; +int num_of_runs;int (*setup)(int num, ...); + };pid_t bm_pid, ppid; +extern char cbm_mask[256 ]; +extern unsigned long long_mask; +extern char llc_occup_path[1024 ];int remount_resctrlfs (bool mum_resctrlfs);int get_resource_id (int cpu_no,int *resource_id);
@@ -76,5 +89,13 @@ void tests_cleanup(void);void mbm_test_cleanup (void );int mba_schemata_change (int cpu_no,char *bw_report,char **benchmark_cmd);void mba_test_cleanup (void ); +int get_cbm_mask (char *cache_type); +int get_cache_size (int cpu_no,char *cache_type,unsigned long *cache_size); +void ctrlc_handler (int signum,siginfo_t *info,void *ptr); +int cqm_resctrl_val (int cpu_no,int n,char **benchmark_cmd); +unsigned int count_bits (unsigned long n); +void cqm_test_cleanup (void ); +int get_core_sibling (int cpu_no); +int measure_cache_vals (struct resctrl_val_param *param,int bm_pid);#endif /* RESCTRL_H */
tools/testing/selftests/resctrl/resctrl_tests.c +46 -10
128 lines not on this page · raw message
tools/testing/selftests/resctrl/resctrl_val.c +58 -37
168 lines not on this page · raw message
tools/testing/selftests/resctrl/resctrlfs.c +211 -16
328 lines not on this page · raw message
Signature · 1 line
2.7.4
Re: [PATCH v7 08/13] selftests/resctrl: Add Cache QoS Monitoring (CQM) selftest
On Sat, 9 Feb 2019 18:50:37 -0800
Fenghua Yu <[email protected]> wrote:
Hi,
Quoted 157 lines
From: Arshiya Hayatkhan Pathan <[email protected]>
Cache QoS Monitoring (CQM) selftest starts stressful cache benchmark with specified size of memory to access the cache. Last Level cache occupancy reported by CQM should be close to the size of the memory.
Signed-off-by Arshiya Hayatkhan Pathan <[email protected]>Signed-off-by Sai Praneeth Prakhya <[email protected]>Signed-off-by Fenghua Yu <[email protected]>Signed-off-by Babu Moger <[email protected]>--- tools/testing/selftests/resctrl/Makefile | 2 +- tools/testing/selftests/resctrl/cache.c | 101 +++++++++++ tools/testing/selftests/resctrl/cqm_test.c | 173 ++++++++++++++++++ tools/testing/selftests/resctrl/mba_test.c | 1 + tools/testing/selftests/resctrl/mbm_test.c | 1 + tools/testing/selftests/resctrl/resctrl.h | 23 ++- tools/testing/selftests/resctrl/resctrl_tests.c | 56 ++++-- tools/testing/selftests/resctrl/resctrl_val.c | 95 ++++++---- tools/testing/selftests/resctrl/resctrlfs.c | 227 ++++++++++++++++++++++-- 9 files changed, 614 insertions(+), 65 deletions(-) create mode 100644 tools/testing/selftests/resctrl/cache.c create mode 100644 tools/testing/selftests/resctrl/cqm_test.ctools/testing/selftests/resctrl/Makefile +1
-1@@ -8,7 +8,7 @@ all: resctrl_tests resctrl_tests: *.o $(CC) $(CFLAGS) -o resctrl_tests resctrl_tests.o resctrlfs.o \ - resctrl_val.o fill_buf.o mbm_test.o mba_test.o + resctrl_val.o fill_buf.o mbm_test.o mba_test.o cache.o cqm_test.o .PHONY: cleantools/testing/selftests/resctrl/cache.c new file +101
-0@@ -0,0 +1,101 @@ +// SPDX-License-Identifier: GPL-2.0 + +#include <stdint.h> +#include "resctrl.h" + +struct read_format { +__u64 nr;/* The number of events */ +struct { +__u64 value;/* The value of the event */ + } values[2 ]; +}; + +char cbm_mask[256 ]; +unsigned long long_mask; +char llc_occup_path[1024 ]; + +/* +* Get LLC Occupancy as reported by RESCTRL FS +* For CQM, +* 1. If con_mon grp and mon grp given, then read from mon grp in +* con_mon grp +* 2. If only con_mon grp given, then read from con_mon grp +* 3. If both not given, then read from root con_mon grp +* For CAT, +* 1. If con_mon grp given, then read from it +* 2. If con_mon grp not given, then read from root con_mon grp +* +* Return: =0 on success. <0 on failure. +*/ +static int get_llc_occu_resctrl (unsigned long *llc_occupancy) +{ +FILE *fp; + + fp =fopen (llc_occup_path,"r" ); +if (!fp) { +perror ("Failed to open results file" ); + +return errno; + } +if (fscanf (fp,"%lu" , llc_occupancy) <=0 ) { +perror ("Could not get llc occupancy" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + +return 0 ; +} + +/* +* print_results_cache: the cache results are stored in a file +* @filename: file that stores the results +* @bm_pid: child pid that runs benchmark +* @llc_value: perf miss value / +* llc occupancy value reported by resctrl FS +* +* Return: 0 on success. non-zero on failure. +*/ +static int print_results_cache (char *filename,int bm_pid, +unsigned long llc_value) +{ +FILE *fp; + +if (strcmp (filename,"stdio" ) ==0 ||strcmp (filename,"stderr" ) ==0 ) { +printf ("Pid: %d \t LLC_value: %lu\n" , bm_pid, + llc_value); + }else { + fp =fopen (filename,"a" ); +if (!fp) { +perror ("Cannot open results file" ); + +return errno; + } +fprintf (fp,"Pid: %d \t llc_value: %lu\n" , bm_pid, llc_value); +fclose (fp); + } + +return 0 ; +} + +int measure_cache_vals (struct resctrl_val_param *param,int bm_pid) +{ +unsigned long llc_occu_resc =0 , llc_value =0 ; +int ret; + +/* +* Measure llc occupancy from resctrl. +*/ +if (!strcmp (param->resctrl_val,"cqm" )) { + ret =get_llc_occu_resctrl (&llc_occu_resc); +if (ret <0 ) +return ret; + llc_value = llc_occu_resc; + } + ret =print_results_cache (param->filename, bm_pid, llc_value); +if (ret) +return ret; + +return 0 ; +}tools/testing/selftests/resctrl/cqm_test.c new file +7
-0@@ -0,0 +1,173 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Cache Monitoring Technology (CQM) test +* +* Copyright (C) 2018 Intel Corporation +* +* Authors:
tools/testing/selftests/resctrl/cqm_test.c line 17
Quoted 10 lines
tools/testing/selftests/resctrl/cqm_test.c new file +10
-0+* Arshiya Hayatkhan Pathan <[email protected]> +* Sai Praneeth Prakhya <[email protected]>, +* Fenghua Yu <[email protected]> +*/ +#include "resctrl.h" +#include <unistd.h> + +#define RESULT_FILE_NAME "result_cqm" +#define NUM_OF_RUNS 5 +#define MAX_DIFF 2000000
What unit is this?
Quoted 19 lines
tools/testing/selftests/resctrl/cqm_test.c new file +19
-0+#define MAX_DIFF_PERCENT 15 + +int count_of_bits; +char cbm_mask[256 ]; +unsigned long long_mask; +unsigned long cache_size; + +static int cqm_setup (int num, ...) +{ +struct resctrl_val_param *p; +va_list param; + +va_start (param, num); + p = va_arg(param, struct resctrl_val_param *); +va_end (param); + +/* Run NUM_OF_RUNS times */ +if (p->num_of_runs >=NUM_OF_RUNS ) +return -1 ;
tools/testing/selftests/resctrl/cqm_test.c line 46
Quoted 10 lines
tools/testing/selftests/resctrl/cqm_test.c new file +10
-0+ + p->num_of_runs++; + +return 0 ; +} + +static void show_cache_info (unsigned long sum_llc_occu_resc,int no_of_bits, +unsigned long span) +{ +unsigned long avg_llc_occu_resc =0 ;
No need to initialise.
tools/testing/selftests/resctrl/cqm_test.c line 48
Quoted 2 lines
tools/testing/selftests/resctrl/cqm_test.c new file +2
-0+long avg_diff =0 ; +float diff_percent;
Why is this a float?
tools/testing/selftests/resctrl/cqm_test.c line 51
Quoted 2 lines
tools/testing/selftests/resctrl/cqm_test.c new file +2
-0+ avg_llc_occu_resc = sum_llc_occu_resc / (NUM_OF_RUNS -1 ); + avg_diff = (long )abs (span - avg_llc_occu_resc);
This looks strange to me. First: abs() is definitely wrong, it should be labs(). But this cast seems pointless, I think you want to have:
avg_diff = labs((long int)(span - avg_llc_occu_resc));
The cast should not be necessary (unsigned - unsigned can be signed), but gcc warns about it without that.
tools/testing/selftests/resctrl/cqm_test.c line 58
Quoted 7 lines
tools/testing/selftests/resctrl/cqm_test.c new file +7
-0+ +printf ("\nResults are displayed in (Bytes)\n" ); +printf ("\nNumber of bits: %d \t" , no_of_bits); +printf ("Avg_llc_occu_resc: %lu \t" , avg_llc_occu_resc); +printf ("llc_occu_exp (span): %lu \t" , span); + + diff_percent = (((float )span - avg_llc_occu_resc) / span) *100 ;
I would just write this as:
diff_percent = avg_diff * 100 / span;
tools/testing/selftests/resctrl/cqm_test.c line 64
Quoted 6 lines
tools/testing/selftests/resctrl/cqm_test.c new file +6
-0+ +printf ("Diff: %ld \t" , avg_diff); +printf ("Percent diff=%d\t" ,abs ((int )diff_percent)); + +if ((abs ((int )diff_percent) <=MAX_DIFF_PERCENT ) || + (abs (avg_diff) <=MAX_DIFF ))
... which allows you to get rid of all those fancy casts and abs() calls here.
tools/testing/selftests/resctrl/cqm_test.c line 72
Quoted 8 lines
tools/testing/selftests/resctrl/cqm_test.c new file +8
-0+printf ("Passed\n" ); + else + printf("Failed\n"); +} + +static int check_results (struct resctrl_val_param *param,int no_of_bits) +{ +char *token_array[8 ], temp[512 ];
This size here does not match ....
tools/testing/selftests/resctrl/cqm_test.c line 85
Quoted 13 lines
tools/testing/selftests/resctrl/cqm_test.c new file +13
-0+unsigned long sum_llc_occu_resc =0 ; +int runs =0 ; +FILE *fp; + +printf ("\nchecking for pass/fail\n" ); + fp =fopen (param->filename,"r" ); +if (!fp) { +perror ("Error in opening file\n" ); + +return errno; + } + +while (fgets (temp,1024 , fp)) {
this user. Just use sizeof(temp).
Cheers,
Andre.
Quoted 838 lines
tools/testing/selftests/resctrl/cqm_test.c new file +88
-0+char *token =strtok (temp,":\t" ); +int fields =0 ; + +while (token) { + token_array[fields++] = token; + token =strtok (NULL ,":\t" ); + } + +/* Field 3 is llc occ resc value */ +if (runs >0 ) + sum_llc_occu_resc +=strtoul (token_array[3 ],NULL ,0 ); + runs++; + } +fclose (fp); +show_cache_info (sum_llc_occu_resc, no_of_bits, param->span); + +return 0 ; +} + +void cqm_test_cleanup (void ) +{ +remove (RESULT_FILE_NAME ); +} + +int cqm_resctrl_val (int cpu_no,int n,char **benchmark_cmd) +{ +int ret, mum_resctrlfs; + + cache_size =0 ; + mum_resctrlfs =1 ; + + ret =remount_resctrlfs (mum_resctrlfs); +if (ret) +return ret; + + ret =validate_resctrl_feature_request ("cqm" ); +if (ret) +return ret; + + ret =get_cbm_mask ("L3" ); +if (ret) +return ret; + + long_mask =strtoul (cbm_mask,NULL ,16 ); + + ret =get_cache_size (cpu_no,"L3" , &cache_size); +if (ret) +return ret; +printf ("cache size :%lu\n" , cache_size); + + count_of_bits =count_bits (long_mask); + +if (n <1 || n > count_of_bits) { +printf ("Invalid input value for numbr_of_bits n!\n" ); +printf ("Please Enter value in range 1 to %d\n" , count_of_bits); +return -1 ; + } + +struct resctrl_val_param param = { + .resctrl_val ="cqm" , + .ctrlgrp ="c1" , + .mongrp ="m1" , + .cpu_no = cpu_no, + .mum_resctrlfs =0 , + .filename =RESULT_FILE_NAME , + .mask = ~(long_mask << n) & long_mask, + .span = cache_size * n / count_of_bits, + .num_of_runs =0 , + .setup = cqm_setup, + }; + +if (strcmp (benchmark_cmd[0 ],"fill_buf" ) ==0 ) +sprintf (benchmark_cmd[1 ],"%llu" , param.span); + +remove (RESULT_FILE_NAME ); + + ret =resctrl_val (benchmark_cmd, ¶m); +if (ret) +return ret; + + ret =check_results (¶m, n); +if (ret) +return ret; + +cqm_test_cleanup (); + +return 0 ; +}tools/testing/selftests/resctrl/mba_test.c +1
-0@@ -57,6 +57,7 @@ static void show_mba_info(unsigned long *bw_imc, unsigned long *bw_resc) {int allocation, failed =0 , runs; +printf ("\nResults are displayed in (MB)\n" );/* Memory bandwidth from 100% down to 10% */ for (allocation =0 ; allocation <ALLOCATION_MAX /ALLOCATION_STEP ; allocation++) {tools/testing/selftests/resctrl/mbm_test.c +1
-0@@ -36,6 +36,7 @@ show_bw_info(unsigned long *bw_imc, unsigned long *bw_resc, int span) avg_bw_resc = sum_bw_resc /4 ; avg_diff = avg_bw_resc - avg_bw_imc; +printf ("\nResults are displayed in (MB)\n" );printf ("\nSpan (MB): %d \t" , span);printf ("avg_bw_imc: %lu\t" , avg_bw_imc);printf ("avg_bw_resc: %lu\t" , avg_bw_resc);tools/testing/selftests/resctrl/resctrl.h +22
-1@@ -16,11 +16,18 @@#include <sys/ioctl.h> #include <sys/mount.h> #include <sys/types.h> +#include <sys/select.h> #include <asm/unistd.h> #include <linux/perf_event.h> +#include <sys/time.h> +#include <math.h> +#include <sys/wait.h> +#include <sys/eventfd.h> +#define MB (1024 * 1024)#define RESCTRL_PATH "/sys/fs/resctrl"#define PHYS_ID_PATH "/sys/devices/system/cpu/cpu" +#define CBM_MASK_PATH "/sys/fs/resctrl/info"#define PARENT_EXIT (err_msg) \ do { \@@ -46,14 +53,20 @@ struct resctrl_val_param {char ctrlgrp[64 ];char mongrp[64 ];int cpu_no; -int span; +unsigned long long span;int mum_resctrlfs;char filename[64 ];char *bw_report; +unsigned long mask; +int num_of_runs;int (*setup)(int num, ...); + };pid_t bm_pid, ppid; +extern char cbm_mask[256 ]; +extern unsigned long long_mask; +extern char llc_occup_path[1024 ];int remount_resctrlfs (bool mum_resctrlfs);int get_resource_id (int cpu_no,int *resource_id);@@ -76,5 +89,13 @@ void tests_cleanup(void);void mbm_test_cleanup (void );int mba_schemata_change (int cpu_no,char *bw_report,char **benchmark_cmd);void mba_test_cleanup (void ); +int get_cbm_mask (char *cache_type); +int get_cache_size (int cpu_no,char *cache_type,unsigned long *cache_size); +void ctrlc_handler (int signum,siginfo_t *info,void *ptr); +int cqm_resctrl_val (int cpu_no,int n,char **benchmark_cmd); +unsigned int count_bits (unsigned long n); +void cqm_test_cleanup (void ); +int get_core_sibling (int cpu_no); +int measure_cache_vals (struct resctrl_val_param *param,int bm_pid);#endif /* RESCTRL_H */ tools/testing/selftests/resctrl/resctrl_tests.c +46
-10@@ -16,11 +16,13 @@static void cmd_help (void ) { -printf ("usage: resctrl_tests [-h] [-b \"benchmark_cmd [options]\"] [-t test list]\n" ); -printf ("\t-b benchmark_cmd [options]: run specified benchmark\n" ); +printf ("usage: resctrl_tests [-h] [-b \"benchmark_cmd [options]\"] [-t test list] [-n no_of_bits]\n" ); +printf ("\t-b benchmark_cmd [options]: run specified benchmark for MBM, MBA and CQM" );printf ("\t default benchmark is builtin fill_buf\n" );printf ("\t-t test list: run tests specified in the test list, " ); -printf ("e.g. -t mbm,mba\n" ); +printf ("e.g. -t mbm, mba, cqm\n" ); +printf ("\t-n no_of_bits: run cache tests using specified no of bits in cache bit mask\n" ); +printf ("\t-p cpu_no: specify CPU number to run the test. 1 is default\n" );printf ("\t-h: help\n" ); }@@ -28,20 +30,22 @@ void tests_cleanup(void) {mbm_test_cleanup ();mba_test_cleanup (); +cqm_test_cleanup (); }int main (int argc,char **argv) {char benchmark_cmd_area[BENCHMARK_ARGS ][BENCHMARK_ARG_SIZE ]; -int res, c, cpu_no =1 , span =250 , argc_new = argc, i; -int ben_ind; -bool has_ben = false, mbm_test = true, mba_test = true; +int res, c, cpu_no =1 , span =250 , argc_new = argc, i, no_of_bits =5 ; +int ben_ind, ben_count; +bool has_ben = false, mbm_test = true, mba_test = true, cqm_test = true;char *benchmark_cmd[BENCHMARK_ARGS ];char bw_report[64 ], bm_type[64 ];for (i =0 ; i < argc; i++) {if (strcmp (argv[i],"-b" ) ==0 ) { ben_ind = i +1 ; + ben_count = argc - ben_ind; argc_new = ben_ind -1 ; has_ben =1 ;break ;@@ -57,11 +61,14 @@ int main(int argc, char **argv) mbm_test = false; mba_test = false; + cqm_test = false;while (token) {if (!strcmp (token,"mbm" )) { mbm_test = true; }else if (!strcmp (token,"mba" )) { mba_test = true; + }else if (!strcmp (token,"cqm" )) { + cqm_test = true; }else {printf ("invalid argument\n" );@@ -70,6 +77,12 @@ int main(int argc, char **argv) token =strtok (NULL ,":\t" ); }break ; +case 'n' : + no_of_bits =atoi (optarg); +break ; +case 'p' : + cpu_no =atoi (optarg); +break ;case 'h' :cmd_help ();@@ -92,16 +105,25 @@ int main(int argc, char **argv)return errno; } -if (!has_ben) { +if (has_ben) { +/* Extract benchmark command from command line. */ +for (i = ben_ind; i < argc; i++) { + benchmark_cmd[i - ben_ind] = benchmark_cmd_area[i]; +sprintf (benchmark_cmd[i - ben_ind],"%s" , argv[i]); + } + benchmark_cmd[ben_count] =NULL ; + }else {/* If no benchmark is given by "-b" argument, use fill_buf. */ -for (i =0 ; i <5 ; i++) +for (i =0 ; i <6 ; i++) benchmark_cmd[i] = benchmark_cmd_area[i]; +strcpy (benchmark_cmd[0 ],"fill_buf" );sprintf (benchmark_cmd[1 ],"%d" , span);strcpy (benchmark_cmd[2 ],"1" );strcpy (benchmark_cmd[3 ],"1" );strcpy (benchmark_cmd[4 ],"0" ); - benchmark_cmd[5 ] =NULL ; +strcpy (benchmark_cmd[5 ],"" ); + benchmark_cmd[6 ] =NULL ; }sprintf (bw_report,"reads" );@@ -109,18 +131,32 @@ int main(int argc, char **argv)if (mbm_test) {printf ("\nMBM BW Change Starting..\n" ); +if (!has_ben) +sprintf (benchmark_cmd[5 ],"%s" ,"mbm" ); res =mbm_bw_change (span, cpu_no, bw_report, benchmark_cmd);if (res)printf ("Error in running tests for mbm bw change!\n" ); +mbm_test_cleanup (); }if (mba_test) {printf ("\nMBA Schemata Change Starting..\n" );if (!has_ben) -sprintf (benchmark_cmd[1 ],"%d" , span); +sprintf (benchmark_cmd[5 ],"%s" ,"mba" ); res =mba_schemata_change (cpu_no, bw_report, benchmark_cmd);if (res)printf ("Error in tests for mba-change-schemata!\n" ); +mba_test_cleanup (); + } + +if (cqm_test) { +printf ("\nCQM Test Starting..\n" ); +if (!has_ben) +sprintf (benchmark_cmd[5 ],"%s" ,"cqm" ); + res =cqm_resctrl_val (cpu_no, no_of_bits, benchmark_cmd); +if (res) +printf ("Error in CQM test!\n" ); +cqm_test_cleanup (); }return 0 ;tools/testing/selftests/resctrl/resctrl_val.c +58
-37@@ -11,7 +11,6 @@*/ #include "resctrl.h" -#define MB (1024 * 1024)#define UNCORE_IMC "uncore_imc"#define READ_FILE_NAME "events/cas_count_read"#define WRITE_FILE_NAME "events/cas_count_write"@@ -33,6 +32,18 @@#define MBM_LOCAL_BYTES_PATH \ "%s/mon_data/mon_L3_%02d/mbm_local_bytes" +#define CON_MON_LCC_OCCUP_PATH \ + "%s/%s/mon_groups/%s/mon_data/mon_L3_%02d/llc_occupancy" + +#define CON_LCC_OCCUP_PATH \ + "%s/%s/mon_data/mon_L3_%02d/llc_occupancy" + +#define MON_LCC_OCCUP_PATH \ + "%s/mon_groups/%s/mon_data/mon_L3_%02d/llc_occupancy" + +#define LCC_OCCUP_PATH \ + "%s/mon_data/mon_L3_%02d/llc_occupancy" +struct membw_read_format {__u64 value;/* The value of the event */ __u64 time_enabled;/* if PERF_FORMAT_TOTAL_TIME_ENABLED */ @@ -207,12 +218,12 @@ static int read_from_imc_dir(char *imc_dir, int count)* A config again has two parts, event and umask. * Enumerate all these details into an array of structures. * -* Return: >= 0 on success. < 0 on failure. +* Return: > 0 on success. <= 0 on failure. */ static int num_of_imcs (void ) {unsigned int count =0 ; -char imc_dir[1024 ]; +char imc_dir[512 ];struct dirent *ep;int ret;DIR *dp;@@ -434,16 +445,6 @@ static unsigned long get_mem_bw_resctrl(void)pid_t bm_pid, ppid; -static void ctrlc_handler (int signum,siginfo_t *info,void *ptr) -{ -kill (bm_pid,SIGKILL ); -umount_resctrlfs (); -tests_cleanup (); -printf ("Ending\n\n" ); - -exit (EXIT_SUCCESS ); -} -/* * print_results_bw: the memory bandwidth results are stored in a file * @filename: file that stores the results @@ -482,6 +483,42 @@ static int print_results_bw(char *filename, int bm_pid, float bw_imc,return 0 ; } +static void set_cqm_path (const char *ctrlgrp,const char *mongrp,char sock_num) +{ +if (strlen (ctrlgrp) &&strlen (mongrp)) +sprintf (llc_occup_path,CON_MON_LCC_OCCUP_PATH ,RESCTRL_PATH , + ctrlgrp, mongrp, sock_num); +else if (!strlen (ctrlgrp) &&strlen (mongrp)) +sprintf (llc_occup_path,MON_LCC_OCCUP_PATH ,RESCTRL_PATH , + mongrp, sock_num); +else if (strlen (ctrlgrp) && !strlen (mongrp)) +sprintf (llc_occup_path,CON_LCC_OCCUP_PATH ,RESCTRL_PATH , + ctrlgrp, sock_num); +else if (!strlen (ctrlgrp) && !strlen (mongrp)) +sprintf (llc_occup_path,LCC_OCCUP_PATH ,RESCTRL_PATH , sock_num); +} + +/* +* initialize_llc_occu_resctrl: Appropriately populate "llc_occup_path" +* @ctrlgrp: Name of the control monitor group (con_mon grp) +* @mongrp: Name of the monitor group (mon grp) +* @cpu_no: CPU number that the benchmark PID is binded to +* @resctrl_val: Resctrl feature (Eg: cat, cqm.. etc) +*/ +static void initialize_llc_occu_resctrl (const char *ctrlgrp,const char *mongrp, +int cpu_no,char *resctrl_val) +{ +int resource_id; + +if (get_resource_id (cpu_no, &resource_id) <0 ) { +perror ("Unable to resource_id" ); +return ; + } + +if (strcmp (resctrl_val,"cqm" ) ==0 ) +set_cqm_path (ctrlgrp, mongrp, resource_id); +} +static int measure_vals (struct resctrl_val_param *param,unsigned long *bw_resc_start) {@@ -528,14 +565,10 @@ int resctrl_val(char **benchmark_cmd, struct resctrl_val_param *param)unsigned long bw_resc_start =0 ;struct sigaction sigact;union sigval value; -FILE *fp;if (strcmp (param->filename,"" ) ==0 )sprintf (param->filename,"stdio" ); -if (strcmp (param->bw_report,"" ) ==0 ) - param->bw_report ="total" ; -if ((strcmp (resctrl_val,"mba" )) ==0 || (strcmp (resctrl_val,"mbm" )) ==0 ) { ret =validate_bw_report_request (param->bw_report);@@ -553,21 +586,6 @@ int resctrl_val(char **benchmark_cmd, struct resctrl_val_param *param)*/ ppid =getpid (); -/* File based synchronization between parent and child */ - fp =fopen ("sig" ,"w" ); -if (!fp) { -perror ("Failed to open sig file" ); - -return -1 ; - } -if (fprintf (fp,"%d\n" ,0 ) <=0 ) { -perror ("Unable to establish sync bw parent & child" ); -fclose (fp); - -return -1 ; - } -fclose (fp); -if (pipe (pipefd)) {perror ("Unable to create pipe" );@@ -649,7 +667,9 @@ int resctrl_val(char **benchmark_cmd, struct resctrl_val_param *param)initialize_mem_bw_resctrl (param->ctrlgrp, param->mongrp, param->cpu_no, resctrl_val); - } + }else if (strcmp (resctrl_val,"cqm" ) ==0 ) +initialize_llc_occu_resctrl (param->ctrlgrp, param->mongrp, + param->cpu_no, resctrl_val);/* Parent waits for child to be ready. */ close (pipefd[1 ]);@@ -669,7 +689,8 @@ int resctrl_val(char **benchmark_cmd, struct resctrl_val_param *param)/* Test runs until the callback setup() tells the test to stop. */ while (1 ) { -if (strcmp (resctrl_val,"mbm" ) ==0 ) { +if ((strcmp (resctrl_val,"mbm" ) ==0 ) || + (strcmp (resctrl_val,"mba" ) ==0 )) { ret = param->setup (1 , param);if (ret) { ret =0 ;@@ -679,14 +700,14 @@ int resctrl_val(char **benchmark_cmd, struct resctrl_val_param *param) ret =measure_vals (param, &bw_resc_start);if (ret)break ; - }else if ((strcmp (resctrl_val,"mba" ) ==0 )) { + }else if (strcmp (resctrl_val,"cqm" ) ==0 ) { ret = param->setup (1 , param);if (ret) { ret =0 ;break ; } - - ret =measure_vals (param, &bw_resc_start); +sleep (1 ); + ret =measure_cache_vals (param, bm_pid);if (ret)break ; }else {tools/testing/selftests/resctrl/resctrlfs.c +211
-16@@ -13,12 +13,18 @@#define RESCTRL_MBM "L3 monitoring detected"#define RESCTRL_MBA "MB allocation detected" +#define RESCTRL_CQM "L3 monitoring detected" +#define CORE_SIBLINGS_PATH "/sys/bus/cpu/devices/cpu" +enum {RESCTRL_FEATURE_MBM ,RESCTRL_FEATURE_MBA , +RESCTRL_FEATURE_CQM ,MAX_RESCTRL_FEATURES }; +char cbm_mask[256 ]; +/* * remount_resctrlfs - Remount resctrl FS at /sys/fs/resctrl * @mum_resctrlfs: Should the resctrl FS be remounted? @@ -62,7 +68,7 @@ int remount_resctrlfs(bool mum_resctrlfs)return errno; } -printf ("Remount: done!\n" ); +printf ("umount: done!\n" ); }else {printf ("Mounted already. Not remounting!\n" );@@ -122,6 +128,143 @@ int get_resource_id(int cpu_no, int *resource_id) }/* +* get_cache_size - Get cache size for a specified CPU +* @cpu_no: CPU number +* @cache_type: Cache level L2/L3 +* @cache_size: pointer to cache_size +* +* Return: = 0 on success, < 0 on failure. +*/ +int get_cache_size (int cpu_no,char *cache_type,unsigned long *cache_size) +{ +char cache_path[1024 ], cache_str[64 ]; +int length, i, cache_num; +FILE *fp; + +if (!strcmp (cache_type,"L3" )) { + cache_num =3 ; + }else if (!strcmp (cache_type,"L2" )) { + cache_num =2 ; + }else { +perror ("Invalid cache level" ); +return -1 ; + } + +sprintf (cache_path,"/sys/bus/cpu/devices/cpu%d/cache/index%d/size" , + cpu_no, cache_num); + fp =fopen (cache_path,"r" ); +if (!fp) { +perror ("Failed to open cache size" ); + +return -1 ; + } +if (fscanf (fp,"%s" , cache_str) <=0 ) { +perror ("Could not get cache_size" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + + length = (int )strlen (cache_str); + + *cache_size =0 ; + +for (i =0 ; i < length; i++) { +if ((cache_str[i] >='0' ) && (cache_str[i] <='9' )) + + *cache_size = *cache_size *10 + (cache_str[i] -'0' ); + +else if (cache_str[i] =='K' ) + + *cache_size = *cache_size *1024 ; + +else if (cache_str[i] =='M' ) + + *cache_size = *cache_size *1024 *1024 ; + +else +break ; + } + +return 0 ; +} + +/* +* get_cbm_mask - Get cbm mask for given cache +* @cache_type: Cache level L2/L3 +* +* Mask is stored in cbm_mask which is global variable. +* +* Return: = 0 on success, < 0 on failure. +*/ +int get_cbm_mask (char *cache_type) +{ +char cbm_mask_path[1024 ]; +FILE *fp; + +sprintf (cbm_mask_path,"%s/%s/cbm_mask" ,CBM_MASK_PATH , cache_type); + + fp =fopen (cbm_mask_path,"r" ); +if (!fp) { +perror ("Failed to open cache level" ); + +return -1 ; + } +if (fscanf (fp,"%s" , cbm_mask) <=0 ) { +perror ("Could not get max cbm_mask" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + +return 0 ; +} + +/* +* get_core_sibling - Get sibling core id from the same socket for given CPU +* @cpu_no: CPU number +* +* Return: > 0 on success, < 0 on failure. +*/ +int get_core_sibling (int cpu_no) +{ +char core_siblings_path[1024 ], cpu_list_str[64 ]; +int sibling_cpu_no =-1 ; +FILE *fp; + +sprintf (core_siblings_path,"%s%d/topology/core_siblings_list" , +CORE_SIBLINGS_PATH , cpu_no); + + fp =fopen (core_siblings_path,"r" ); +if (!fp) { +perror ("Failed to open core siblings path" ); + +return -1 ; + } +if (fscanf (fp,"%s" , cpu_list_str) <=0 ) { +perror ("Could not get core_siblings list" ); +fclose (fp); + +return -1 ; + } +fclose (fp); + +char *token =strtok (cpu_list_str,"-," ); + +while (token) { + sibling_cpu_no =atoi (token); +/* Skipping core 0 as we don't want to run test on core 0 */ +if (sibling_cpu_no !=0 ) +break ; + token =strtok (NULL ,"-," ); + } + +return sibling_cpu_no; +} + +/* * taskset_benchmark - Taskset PID (i.e. benchmark) to a specified cpu * @bm_pid: PID that should be binded * @cpu_no: CPU number at which the PID would be binded @@ -157,9 +300,10 @@ int taskset_benchmark(pid_t bm_pid, int cpu_no)*/ void run_benchmark (int signum,siginfo_t *info,void *ucontext) { -unsigned long long span; -int operation, ret; +int operation, ret, malloc_and_init_memory, memflush; +unsigned long long span, buffer_span;char **benchmark_cmd; +char resctrl_val[64 ];FILE *fp; benchmark_cmd = info->si_ptr;@@ -175,8 +319,18 @@ void run_benchmark(int signum, siginfo_t *info, void *ucontext)if (strcmp (benchmark_cmd[0 ],"fill_buf" ) ==0 ) {/* Execute default fill_buf benchmark */ span =strtoul (benchmark_cmd[1 ],NULL ,10 ); + malloc_and_init_memory =atoi (benchmark_cmd[2 ]); + memflush =atoi (benchmark_cmd[3 ]); operation =atoi (benchmark_cmd[4 ]); -if (run_fill_buf (span,1 ,1 , operation,NULL )) +sprintf (resctrl_val,"%s" , benchmark_cmd[5 ]); + +if (strcmp (resctrl_val,"cqm" ) !=0 ) + buffer_span = span *MB ; +else + buffer_span = span; + +if (run_fill_buf (buffer_span, malloc_and_init_memory, memflush, + operation, resctrl_val))fprintf (stderr,"Error in running fill buffer\n" ); }else {/* Execute specified benchmark */ @@ -203,6 +357,14 @@ static int create_grp(const char *grp_name, char *grp, const char *parent_grp)struct dirent *ep;DIR *dp; +/* +* At this point, we are guaranteed to have resctrl FS mounted and if +* length of grp_name == 0, it means, user wants to use root con_mon +* grp, so do nothing +*/ +if (strlen (grp_name) ==0 ) +return 0 ; +/* Check if requested grp exists or not */ dp =opendir (parent_grp);if (dp) {@@ -268,11 +430,11 @@ static int write_pid_to_tasks(char *tasks, pid_t pid)int write_bm_pid_to_resctrl (pid_t bm_pid,char *ctrlgrp,char *mongrp,char *resctrl_val) { -char controlgroup[256 ], monitorgroup[256 ], monitorgroup_p[256 ]; -char tasks[256 ]; +char controlgroup[128 ], monitorgroup[512 ], monitorgroup_p[256 ]; +char tasks[1024 ];int ret; -if (ctrlgrp) +if (strlen (ctrlgrp))sprintf (controlgroup,"%s/%s" ,RESCTRL_PATH , ctrlgrp);else sprintf (controlgroup,"%s" ,RESCTRL_PATH );@@ -286,9 +448,10 @@ int write_bm_pid_to_resctrl(pid_t bm_pid, char *ctrlgrp, char *mongrp,if (ret)return ret; -/* Create mon grp and write pid into it for "mbm" test */ -if ((strcmp (resctrl_val,"mbm" ) ==0 )) { -if (mongrp) { +/* Create mon grp and write pid into it for "mbm" and "cqm" test */ +if ((strcmp (resctrl_val,"cqm" ) ==0 ) || + (strcmp (resctrl_val,"mbm" ) ==0 )) { +if (strlen (mongrp)) {sprintf (monitorgroup_p,"%s/mon_groups" , controlgroup);sprintf (monitorgroup,"%s/%s" , monitorgroup_p, mongrp); ret =create_grp (mongrp, monitorgroup, monitorgroup_p);@@ -326,7 +489,8 @@ int write_schemata(char *ctrlgrp, char *schemata, int cpu_no, char *resctrl_val)int resource_id;FILE *fp; -if (strcmp (resctrl_val,"mba" ) ==0 ) { +if ((strcmp (resctrl_val,"mba" ) ==0 ) || + (strcmp (resctrl_val,"cqm" ) ==0 )) {if (!schemata) {printf ("Schemata empty, so not updating\n" );@@ -337,12 +501,18 @@ int write_schemata(char *ctrlgrp, char *schemata, int cpu_no, char *resctrl_val)return -1 ; } -if (ctrlgrp) +if (strlen (ctrlgrp) !=0 )sprintf (controlgroup,"%s/%s/schemata" ,RESCTRL_PATH , ctrlgrp);else sprintf (controlgroup,"%s/schemata" ,RESCTRL_PATH ); -sprintf (schema,"%s%d%c%s" ,"MB:" , resource_id,'=' , schemata); + +if (!strcmp (resctrl_val,"cqm" )) +sprintf (schema,"%s%d%c%s" ,"L3:" , resource_id,'=' , + schemata); +if (strcmp (resctrl_val,"mba" ) ==0 ) +sprintf (schema,"%s%d%c%s" ,"MB:" , resource_id,'=' , + schemata); fp =fopen (controlgroup,"w" );if (!fp) {@@ -359,7 +529,7 @@ int write_schemata(char *ctrlgrp, char *schemata, int cpu_no, char *resctrl_val) }fclose (fp); -printf ("Write schemata to resctrl FS: done!\n" ); +printf ("Write schemata with %s to resctrl FS: done!\n" , schema); }return 0 ;@@ -375,7 +545,7 @@ int validate_resctrl_feature_request(char *resctrl_val) {int resctrl_features_supported[MAX_RESCTRL_FEATURES ];const char *resctrl_features_list[MAX_RESCTRL_FEATURES ] = { -"mbm" ,"mba" }; +"mbm" ,"mba" ,"cqm" };int i, valid_resctrl_feature =-1 ;char line[1024 ];FILE *fp;@@ -419,6 +589,9 @@ int validate_resctrl_feature_request(char *resctrl_val) resctrl_features_supported[RESCTRL_FEATURE_MBM ] =1 ;if ((strstr (line,RESCTRL_MBA )) !=NULL ) resctrl_features_supported[RESCTRL_FEATURE_MBA ] =1 ; +if ((strstr (line,RESCTRL_CQM )) !=NULL ) + resctrl_features_supported[RESCTRL_FEATURE_CQM ] =1 ; + }fclose (fp);@@ -427,7 +600,7 @@ int validate_resctrl_feature_request(char *resctrl_val)/* Is the resctrl feature request supported? */ if (!resctrl_features_supported[valid_resctrl_feature]) { -fprintf (stderr,"resctrl feature not supported!" ); +fprintf (stderr,"resctrl feature not supported!\n" );return -1 ; }@@ -462,3 +635,25 @@ int perf_event_open(struct perf_event_attr *hw_event, pid_t pid, int cpu, group_fd, flags);return ret; } + +void ctrlc_handler (int signum,siginfo_t *info,void *ptr) +{ +kill (bm_pid,SIGKILL ); +umount_resctrlfs (); +tests_cleanup (); +printf ("Ending\n\n" ); + +exit (EXIT_SUCCESS ); +} + +unsigned int count_bits (unsigned long n) +{ +unsigned int count =0 ; + +while (n) { + count += n &1 ; + n >>=1 ; + } + +return count; +}
PATCH v7 9/13 selftests/resctrl: Add Cache Allocation Technology (CAT) selftest
In reply to Fenghua YuFrom: Arshiya Hayatkhan Pathan <[email protected]>
Cache Allocation Technology (CAT) selftest allocates a portion of last level cache and starts a benchmark to read each cache line in this portion of cache. Measure the cache misses in perf and the misses should be equal to the number of cache lines in this portion of cache.
We don't use CQM to calculate cache usage because some CAT enabled platforms don't have CQM.
--- tools/testing/selftests/resctrl/Makefile | 2 +- tools/testing/selftests/resctrl/cache.c | 175 ++++++++++++++++- tools/testing/selftests/resctrl/cat_test.c | 242 ++++++++++++++++++++++++ tools/testing/selftests/resctrl/fill_buf.c | 10 +- tools/testing/selftests/resctrl/resctrl.h | 3 + tools/testing/selftests/resctrl/resctrl_tests.c | 14 +- tools/testing/selftests/resctrl/resctrlfs.c | 10 +- 7 files changed, 448 insertions(+), 8 deletions(-) create mode 100644 tools/testing/selftests/resctrl/cat_test.c
tools/testing/selftests/resctrl/Makefile +1 -1
@@ -8,7 +8,7 @@ all: resctrl_tests resctrl_tests: *.o $(CC) $(CFLAGS) -o resctrl_tests resctrl_tests.o resctrlfs.o \ - resctrl_val.o fill_buf.o mbm_test.o mba_test.o cache.o cqm_test.o + resctrl_val.o fill_buf.o mbm_test.o mba_test.o cache.o cqm_test.o cat_test.o .PHONY: clean
tools/testing/selftests/resctrl/cache.c +174 -1
197 lines not on this page · raw message
tools/testing/selftests/resctrl/cat_test.c new file +242 -0
242 lines not on this page · raw message
tools/testing/selftests/resctrl/fill_buf.c +8 -2
@@ -110,8 +110,11 @@ static int fill_cache_read(unsigned char *start_ptr, unsigned char *end_ptr,int ret =0 ;FILE *fp; -while (1 ) +while (1 ) { ret =fill_one_span_read (start_ptr, end_ptr); +if (!strcmp (resctrl_val,"cat" )) +break ; + }/* Consume read result so that reading memory is not optimized out. */ fp =fopen ("/dev/null" ,"w" );
@@ -126,8 +129,11 @@ static int fill_cache_read(unsigned char *start_ptr, unsigned char *end_ptr,static int fill_cache_write (unsigned char *start_ptr,unsigned char *end_ptr,char *resctrl_val) { -while (1 ) +while (1 ) {fill_one_span_write (start_ptr, end_ptr); +if (!strcmp (resctrl_val,"cat" )) +break ; + }return 0 ; }
tools/testing/selftests/resctrl/resctrl.h +3 -0
@@ -92,6 +92,9 @@ void mba_test_cleanup(void);int get_cbm_mask (char *cache_type);int get_cache_size (int cpu_no,char *cache_type,unsigned long *cache_size);void ctrlc_handler (int signum,siginfo_t *info,void *ptr); +int cat_val (struct resctrl_val_param *param); +void cat_test_cleanup (void ); +int cat_perf_miss_val (int cpu_no,int no_of_bits,char *cache_type);int cqm_resctrl_val (int cpu_no,int n,char **benchmark_cmd);unsigned int count_bits (unsigned long n);void cqm_test_cleanup (void );
tools/testing/selftests/resctrl/resctrl_tests.c +13 -1
@@ -20,7 +20,7 @@ static void cmd_help(void)printf ("\t-b benchmark_cmd [options]: run specified benchmark for MBM, MBA and CQM" );printf ("\t default benchmark is builtin fill_buf\n" );printf ("\t-t test list: run tests specified in the test list, " ); -printf ("e.g. -t mbm, mba, cqm\n" ); +printf ("e.g. -t mbm, mba, cqm, cat\n" );printf ("\t-n no_of_bits: run cache tests using specified no of bits in cache bit mask\n" );printf ("\t-p cpu_no: specify CPU number to run the test. 1 is default\n" );printf ("\t-h: help\n" );
@@ -31,6 +31,7 @@ void tests_cleanup(void)mbm_test_cleanup ();mba_test_cleanup ();cqm_test_cleanup (); +cat_test_cleanup (); }int main (int argc,char **argv)
@@ -39,6 +40,7 @@ int main(int argc, char **argv)int res, c, cpu_no =1 , span =250 , argc_new = argc, i, no_of_bits =5 ;int ben_ind, ben_count;bool has_ben = false, mbm_test = true, mba_test = true, cqm_test = true; +bool cat_test = true;char *benchmark_cmd[BENCHMARK_ARGS ];char bw_report[64 ], bm_type[64 ];
@@ -62,6 +64,7 @@ int main(int argc, char **argv) mbm_test = false; mba_test = false; cqm_test = false; + cat_test = false;while (token) {if (!strcmp (token,"mbm" )) { mbm_test = true;
@@ -69,6 +72,8 @@ int main(int argc, char **argv) mba_test = true; }else if (!strcmp (token,"cqm" )) { cqm_test = true; + }else if (!strcmp (token,"cat" )) { + cat_test = true; }else {printf ("invalid argument\n" );
@@ -158,6 +163,13 @@ int main(int argc, char **argv)printf ("Error in CQM test!\n" );cqm_test_cleanup (); } +if (cat_test) { +printf ("\nCAT Test Starting..\n" ); + res =cat_perf_miss_val (cpu_no, no_of_bits,"L3" ); +if (res) +printf ("Error in CAT test!\n" ); +cat_test_cleanup (); + }return 0 ; }
tools/testing/selftests/resctrl/resctrlfs.c +7 -3
@@ -14,12 +14,14 @@#define RESCTRL_MBM "L3 monitoring detected"#define RESCTRL_MBA "MB allocation detected"#define RESCTRL_CQM "L3 monitoring detected" +#define RESCTRL_L3_CAT "L3 allocation detected"#define CORE_SIBLINGS_PATH "/sys/bus/cpu/devices/cpu"enum {RESCTRL_FEATURE_MBM ,RESCTRL_FEATURE_MBA ,RESCTRL_FEATURE_CQM , +RESCTRL_FEATURE_L3_CAT ,MAX_RESCTRL_FEATURES };
@@ -490,6 +492,7 @@ int write_schemata(char *ctrlgrp, char *schemata, int cpu_no, char *resctrl_val)FILE *fp;if ((strcmp (resctrl_val,"mba" ) ==0 ) || + (strcmp (resctrl_val,"cat" ) ==0 ) || (strcmp (resctrl_val,"cqm" ) ==0 )) {if (!schemata) {printf ("Schemata empty, so not updating\n" );
@@ -507,7 +510,7 @@ int write_schemata(char *ctrlgrp, char *schemata, int cpu_no, char *resctrl_val)else sprintf (controlgroup,"%s/schemata" ,RESCTRL_PATH ); -if (!strcmp (resctrl_val,"cqm" )) +if (!strcmp (resctrl_val,"cat" ) || !strcmp (resctrl_val,"cqm" ))sprintf (schema,"%s%d%c%s" ,"L3:" , resource_id,'=' , schemata);if (strcmp (resctrl_val,"mba" ) ==0 )
@@ -545,7 +548,7 @@ int validate_resctrl_feature_request(char *resctrl_val) {int resctrl_features_supported[MAX_RESCTRL_FEATURES ];const char *resctrl_features_list[MAX_RESCTRL_FEATURES ] = { -"mbm" ,"mba" ,"cqm" }; +"mbm" ,"mba" ,"cat" ,"cqm" };int i, valid_resctrl_feature =-1 ;char line[1024 ];FILE *fp;
@@ -591,7 +594,8 @@ int validate_resctrl_feature_request(char *resctrl_val) resctrl_features_supported[RESCTRL_FEATURE_MBA ] =1 ;if ((strstr (line,RESCTRL_CQM )) !=NULL ) resctrl_features_supported[RESCTRL_FEATURE_CQM ] =1 ; - +if ((strstr (line,RESCTRL_L3_CAT )) !=NULL ) + resctrl_features_supported[RESCTRL_FEATURE_L3_CAT ] =1 ; }fclose (fp);
Signature · 1 line
2.7.4
Re: [PATCH v7 09/13] selftests/resctrl: Add Cache Allocation Technology (CAT) selftest
On Sat, 9 Feb 2019 18:50:38 -0800
Fenghua Yu <[email protected]> wrote:
Hi,
Quoted 299 lines
From: Arshiya Hayatkhan Pathan <[email protected]>
Cache Allocation Technology (CAT) selftest allocates a portion of last level cache and starts a benchmark to read each cache line in this portion of cache. Measure the cache misses in perf and the misses should be equal to the number of cache lines in this portion of cache.
We don't use CQM to calculate cache usage because some CAT enabled platforms don't have CQM.
Signed-off-by Arshiya Hayatkhan Pathan <[email protected]>Signed-off-by Sai Praneeth Prakhya <[email protected]>Signed-off-by Fenghua Yu <[email protected]>Signed-off-by Babu Moger <[email protected]>--- tools/testing/selftests/resctrl/Makefile | 2 +- tools/testing/selftests/resctrl/cache.c | 175 ++++++++++++++++- tools/testing/selftests/resctrl/cat_test.c | 242 ++++++++++++++++++++++++ tools/testing/selftests/resctrl/fill_buf.c | 10 +- tools/testing/selftests/resctrl/resctrl.h | 3 + tools/testing/selftests/resctrl/resctrl_tests.c | 14 +- tools/testing/selftests/resctrl/resctrlfs.c | 10 +- 7 files changed, 448 insertions(+), 8 deletions(-) create mode 100644 tools/testing/selftests/resctrl/cat_test.ctools/testing/selftests/resctrl/Makefile +1
-1@@ -8,7 +8,7 @@ all: resctrl_tests resctrl_tests: *.o $(CC) $(CFLAGS) -o resctrl_tests resctrl_tests.o resctrlfs.o \ - resctrl_val.o fill_buf.o mbm_test.o mba_test.o cache.o cqm_test.o + resctrl_val.o fill_buf.o mbm_test.o mba_test.o cache.o cqm_test.o cat_test.o .PHONY: cleantools/testing/selftests/resctrl/cache.c +174
-1@@ -10,10 +10,107 @@ struct read_format { } values[2 ]; }; +static struct perf_event_attr pea_llc_miss; +static struct read_format rf_cqm; +static int fd_lm;char cbm_mask[256 ];unsigned long long_mask;char llc_occup_path[1024 ]; +static void initialize_perf_event_attr (void ) +{ + pea_llc_miss.type =PERF_TYPE_HARDWARE ; + pea_llc_miss.size =sizeof (struct perf_event_attr ); + pea_llc_miss.read_format =PERF_FORMAT_GROUP ; + pea_llc_miss.exclude_kernel =1 ; + pea_llc_miss.exclude_hv =1 ; + pea_llc_miss.exclude_idle =1 ; + pea_llc_miss.exclude_callchain_kernel =1 ; + pea_llc_miss.inherit =1 ; + pea_llc_miss.exclude_guest =1 ; + pea_llc_miss.disabled =1 ; +} + +static void ioctl_perf_event_ioc_reset_enable (void ) +{ +ioctl (fd_lm,PERF_EVENT_IOC_RESET ,0 ); +ioctl (fd_lm,PERF_EVENT_IOC_ENABLE ,0 ); +} + +static int perf_event_open_llc_miss (pid_t pid,int cpu_no) +{ + fd_lm =perf_event_open (&pea_llc_miss, pid, cpu_no,-1 , +PERF_FLAG_FD_CLOEXEC ); +if (fd_lm ==-1 ) { +perror ("Error opening leader" ); +ctrlc_handler (0 ,NULL ,NULL ); +return -1 ; + } + +return 0 ; +} + +static int initialize_llc_perf (void ) +{ +memset (&pea_llc_miss,0 ,sizeof (struct perf_event_attr )); +memset (&rf_cqm,0 ,sizeof (struct read_format )); + +/* Initialize perf_event_attr structures for HW_CACHE_MISSES */ +initialize_perf_event_attr (); + + pea_llc_miss.config =PERF_COUNT_HW_CACHE_MISSES ; + + rf_cqm.nr =1 ; + +return 0 ; +} + +static int reset_enable_llc_perf (pid_t pid,int cpu_no) +{ +int ret =0 ; + + ret =perf_event_open_llc_miss (pid, cpu_no); +if (ret <0 ) +return ret; + +/* Start counters to log values */ +ioctl_perf_event_ioc_reset_enable (); + +return 0 ; +} + +/* +* get_llc_perf: llc cache miss through perf events +* @cpu_no: CPU number that the benchmark PID is binded to +* +* Perf events like HW_CACHE_MISSES could be used to validate number of +* cache lines allocated. +* +* Return: =0 on success. <0 on failure. +*/ +static int get_llc_perf (unsigned long *llc_perf_miss) +{ +__u64 total_misses; + +/* Stop counters after one span to get miss rate */ + +ioctl (fd_lm,PERF_EVENT_IOC_DISABLE ,0 ); + +if (read (fd_lm, &rf_cqm,sizeof (struct read_format )) ==-1 ) { +perror ("Could not get llc misses through perf" ); + +return -1 ; + } + + total_misses = rf_cqm.values[0 ].value; + +close (fd_lm); + + *llc_perf_miss = total_misses; + +return 0 ; +} +/* * Get LLC Occupancy as reported by RESCTRL FS * For CQM, @@ -81,10 +178,20 @@ static int print_results_cache(char *filename, int bm_pid,int measure_cache_vals (struct resctrl_val_param *param,int bm_pid) { -unsigned long llc_occu_resc =0 , llc_value =0 ; +unsigned long llc_perf_miss =0 , llc_occu_resc =0 , llc_value =0 ;int ret;/* +* Measure cache miss from perf. +*/ +if (!strcmp (param->resctrl_val,"cat" )) { + ret =get_llc_perf (&llc_perf_miss); +if (ret <0 ) +return ret; + llc_value = llc_perf_miss; + } + +/* * Measure llc occupancy from resctrl. */ if (!strcmp (param->resctrl_val,"cqm" )) {@@ -99,3 +206,69 @@ int measure_cache_vals(struct resctrl_val_param *param, int bm_pid)return 0 ; } + +/* +* cache_val: execute benchmark and measure LLC occupancy resctrl +* and perf cache miss for the benchmark +* @param: parameters passed to cache_val() +* +* Return: 0 on success. non-zero on failure. +*/ +int cat_val (struct resctrl_val_param *param) +{ +int malloc_and_init_memory =1 , memflush =1 , operation =0 , ret =0 ; +char *resctrl_val = param->resctrl_val; +pid_t bm_pid; + +if (strcmp (param->filename,"" ) ==0 ) +sprintf (param->filename,"stdio" ); + + bm_pid =getpid (); + +/* Taskset benchmark to specified cpu */ + ret =taskset_benchmark (bm_pid, param->cpu_no); +if (ret) +return ret; + +/* Write benchmark to specified con_mon grp, mon_grp in resctrl FS*/ + ret =write_bm_pid_to_resctrl (bm_pid, param->ctrlgrp, param->mongrp, + resctrl_val); +if (ret) +return ret; + +if ((strcmp (resctrl_val,"cat" ) ==0 )) { + ret =initialize_llc_perf (); +if (ret) +return ret; + } + +/* Test runs until the callback setup() tells the test to stop. */ +while (1 ) { +if (strcmp (resctrl_val,"cat" ) ==0 ) { + ret = param->setup (1 , param); +if (ret) { + ret =0 ; +break ; + } + ret =reset_enable_llc_perf (bm_pid, param->cpu_no); +if (ret) +break ; + +if (run_fill_buf (param->span, malloc_and_init_memory, + memflush, operation, resctrl_val)) { +fprintf (stderr,"Error-running fill buffer\n" ); + ret =-1 ; +break ; + } + +sleep (1 ); + ret =measure_cache_vals (param, bm_pid); +if (ret) +break ; + }else { +break ; + } + } + +return ret; +}tools/testing/selftests/resctrl/cat_test.c new file +50
-0@@ -0,0 +1,242 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Cache Allocation Technology (CAT) test +* +* Copyright (C) 2018 Intel Corporation +* +* Authors: +* Arshiya Hayatkhan Pathan <[email protected]> +* Sai Praneeth Prakhya <[email protected]>, +* Fenghua Yu <[email protected]> +*/ +#include "resctrl.h" +#include <unistd.h> + +#define RESULT_FILE_NAME1 "result_cat1" +#define RESULT_FILE_NAME2 "result_cat2" +#define NUM_OF_RUNS 5 +#define MAX_DIFF_PERCENT 4 +#define MAX_DIFF 1000000 + +int count_of_bits; +char cbm_mask[256 ]; +unsigned long long_mask; +unsigned long cache_size; + +/* +* Change schemata. Write schemata to specified +* con_mon grp, mon_grp in resctrl FS. +* Run 5 times in order to get average values. +*/ +static int cat_setup (int num, ...) +{ +struct resctrl_val_param *p; +char schemata[64 ]; +va_list param; +int ret =0 ; + +va_start (param, num); + p = va_arg(param, struct resctrl_val_param *); +va_end (param); + +/* Run NUM_OF_RUNS times */ +if (p->num_of_runs >=NUM_OF_RUNS ) +return -1 ; + +if (p->num_of_runs ==0 ) { +sprintf (schemata,"%lx" , p->mask); + ret =write_schemata (p->ctrlgrp, schemata, p->cpu_no, + p->resctrl_val); + }
tools/testing/selftests/resctrl/cat_test.c line 60
Quoted 10 lines
tools/testing/selftests/resctrl/cat_test.c new file +10
-0+ p->num_of_runs++; + +return ret; +} + +static void show_cache_info (unsigned long sum_llc_perf_miss,int no_of_bits, +unsigned long span) +{ +unsigned long allocated_cache_lines = span /64 ; +unsigned long avg_llc_perf_miss =0 ;
No need to initialise.
tools/testing/selftests/resctrl/cat_test.c line 61
Quoted 1 line
tools/testing/selftests/resctrl/cat_test.c new file +1
-0+float diff_percent;
Please not floats for no reason. Should be an int, saves you some nasty casts below.
tools/testing/selftests/resctrl/cat_test.c line 65
Quoted 4 lines
tools/testing/selftests/resctrl/cat_test.c new file +4
-0+ + avg_llc_perf_miss = sum_llc_perf_miss / (NUM_OF_RUNS -1 ); + diff_percent = ((float )allocated_cache_lines - avg_llc_perf_miss) / + allocated_cache_lines *100 ;
That looks confusing. Shouldn't it be:
diff_percent = avg_llc_perf_miss * 100 / allocated_cache_lines;
Otherwise I have a hard time to see how this matches the check below.
Quoted 5 lines
tools/testing/selftests/resctrl/cat_test.c new file +5
-0+printf ("Results are displayed in (Bytes)\n" ); +printf ("\nNumber of bits: %d \t" , no_of_bits); +printf ("Avg_llc_perf_miss: %lu \t" , avg_llc_perf_miss); +printf ("Allocated cache lines: %lu \t" , allocated_cache_lines); +printf ("Percent diff=%d \t" ,abs ((int )diff_percent));
tools/testing/selftests/resctrl/cat_test.c line 80
Quoted 10 lines
tools/testing/selftests/resctrl/cat_test.c new file +10
-0+ +if (abs ((int )diff_percent) >MAX_DIFF_PERCENT ) +printf ("Failed\n" ); +else +printf ("Passed\n" ); +} + +static int check_results (struct resctrl_val_param *param) +{ +char *token_array[8 ], temp[512 ];
The size of temp[] here does not match ...
tools/testing/selftests/resctrl/cat_test.c line 93
Quoted 13 lines
tools/testing/selftests/resctrl/cat_test.c new file +13
-0+unsigned long sum_llc_perf_miss =0 ; +int runs =0 , no_of_bits =0 ; +FILE *fp; + +printf ("\nChecking for pass/fail\n" ); + fp =fopen (param->filename,"r" ); +if (!fp) { +perror ("Error in opening file\n" ); + +return errno; + } + +while (fgets (temp,1024 , fp)) {
... the user here. sizeof(temp) for the rescue!
Cheers,
Andre.
Quoted 308 lines
tools/testing/selftests/resctrl/cat_test.c new file +149
-0+char *token =strtok (temp,":\t" ); +int fields =0 ; + +while (token) { + token_array[fields++] = token; + token =strtok (NULL ,":\t" ); + } +/* +* Discard the first value which is inaccurate due to monitoring +* setup transition phase. +*/ +if (runs >0 ) + sum_llc_perf_miss +=strtoul (token_array[3 ],NULL ,0 ); + runs++; + } + +fclose (fp); + no_of_bits =count_bits (param->mask); + +show_cache_info (sum_llc_perf_miss, no_of_bits, param->span); + +return 0 ; +} + +void cat_test_cleanup (void ) +{ +remove (RESULT_FILE_NAME1 ); +remove (RESULT_FILE_NAME2 ); +} + +int cat_perf_miss_val (int cpu_no,int n,char *cache_type) +{ +unsigned long l_mask, l_mask_1; +int ret, pipefd[2 ], pipe_message, mum_resctrlfs, sibling_cpu_no; +pid_t bm_pid; + + cache_size =0 ; + mum_resctrlfs =1 ; + + ret =remount_resctrlfs (mum_resctrlfs); +if (ret) +return ret; + + ret =validate_resctrl_feature_request ("cat" ); +if (ret) +return ret; + +/* Get default cbm mask for L3/L2 cache */ + ret =get_cbm_mask (cache_type); +if (ret) +return ret; + + long_mask =strtoul (cbm_mask,NULL ,16 ); + +/* Get L3/L2 cache size */ + ret =get_cache_size (cpu_no, cache_type, &cache_size); +if (ret) +return ret; +printf ("cache size :%lu\n" , cache_size); + +/* Get max number of bits from default-cabm mask */ + count_of_bits =count_bits (long_mask); + +if (n <1 || n > count_of_bits -1 ) { +printf ("Invalid input value for no_of_bits n!\n" ); +printf ("Please Enter value in range 1 to %d\n" , + count_of_bits -1 ); +return -1 ; + } + +/* Get core id from same socket for running another thread */ + sibling_cpu_no =get_core_sibling (cpu_no); +if (sibling_cpu_no <0 ) +return -1 ; + +struct resctrl_val_param param = { + .resctrl_val ="cat" , + .cpu_no = cpu_no, + .mum_resctrlfs =0 , + .setup = cat_setup, + }; + + l_mask = long_mask >> n; + l_mask_1 = ~l_mask & long_mask; + +/* Set param values for parent thread which will be allocated bitmask +* with (max_bits - n) bits +*/ + param.span = cache_size * (count_of_bits - n) / count_of_bits; +strcpy (param.ctrlgrp,"c2" ); +strcpy (param.mongrp,"m2" ); +strcpy (param.filename,RESULT_FILE_NAME2 ); + param.mask = l_mask; + param.num_of_runs =0 ; + +if (pipe (pipefd)) { +perror ("Unable to create pipe" ); +return -1 ; + } + + bm_pid =fork (); + +/* Set param values for child thread which will be allocated bitmask +* with n bits +*/ +if (bm_pid ==0 ) { + param.mask = l_mask_1; +strcpy (param.ctrlgrp,"c1" ); +strcpy (param.mongrp,"m1" ); + param.span = cache_size * n / count_of_bits; +strcpy (param.filename,RESULT_FILE_NAME1 ); + param.num_of_runs =0 ; + param.cpu_no = sibling_cpu_no; + } + +remove (param.filename); + + ret =cat_val (¶m); +if (ret) +return ret; + + ret =check_results (¶m); +if (ret) +return ret; + +if (bm_pid ==0 ) { +/* Tell parent that child is ready */ +close (pipefd[0 ]); + pipe_message =1 ; +write (pipefd[1 ], &pipe_message,sizeof (pipe_message)); +close (pipefd[1 ]); +while (1 ) + ; + }else { +/* Parent waits for child to be ready. */ +close (pipefd[1 ]); + pipe_message =0 ; +while (pipe_message !=1 ) +read (pipefd[0 ], &pipe_message,sizeof (pipe_message)); +close (pipefd[0 ]); +kill (bm_pid,SIGKILL ); + } + +cat_test_cleanup (); +if (bm_pid) +umount_resctrlfs (); + +return 0 ; +}tools/testing/selftests/resctrl/fill_buf.c +8
-2@@ -110,8 +110,11 @@ static int fill_cache_read(unsigned char *start_ptr, unsigned char *end_ptr,int ret =0 ;FILE *fp; -while (1 ) +while (1 ) { ret =fill_one_span_read (start_ptr, end_ptr); +if (!strcmp (resctrl_val,"cat" )) +break ; + }/* Consume read result so that reading memory is not optimized out. */ fp =fopen ("/dev/null" ,"w" );@@ -126,8 +129,11 @@ static int fill_cache_read(unsigned char *start_ptr, unsigned char *end_ptr,static int fill_cache_write (unsigned char *start_ptr,unsigned char *end_ptr,char *resctrl_val) { -while (1 ) +while (1 ) {fill_one_span_write (start_ptr, end_ptr); +if (!strcmp (resctrl_val,"cat" )) +break ; + }return 0 ; }tools/testing/selftests/resctrl/resctrl.h +3
-0@@ -92,6 +92,9 @@ void mba_test_cleanup(void);int get_cbm_mask (char *cache_type);int get_cache_size (int cpu_no,char *cache_type,unsigned long *cache_size);void ctrlc_handler (int signum,siginfo_t *info,void *ptr); +int cat_val (struct resctrl_val_param *param); +void cat_test_cleanup (void ); +int cat_perf_miss_val (int cpu_no,int no_of_bits,char *cache_type);int cqm_resctrl_val (int cpu_no,int n,char **benchmark_cmd);unsigned int count_bits (unsigned long n);void cqm_test_cleanup (void );tools/testing/selftests/resctrl/resctrl_tests.c +13
-1@@ -20,7 +20,7 @@ static void cmd_help(void)printf ("\t-b benchmark_cmd [options]: run specified benchmark for MBM, MBA and CQM" );printf ("\t default benchmark is builtin fill_buf\n" );printf ("\t-t test list: run tests specified in the test list, " ); -printf ("e.g. -t mbm, mba, cqm\n" ); +printf ("e.g. -t mbm, mba, cqm, cat\n" );printf ("\t-n no_of_bits: run cache tests using specified no of bits in cache bit mask\n" );printf ("\t-p cpu_no: specify CPU number to run the test. 1 is default\n" );printf ("\t-h: help\n" );@@ -31,6 +31,7 @@ void tests_cleanup(void)mbm_test_cleanup ();mba_test_cleanup ();cqm_test_cleanup (); +cat_test_cleanup (); }int main (int argc,char **argv)@@ -39,6 +40,7 @@ int main(int argc, char **argv)int res, c, cpu_no =1 , span =250 , argc_new = argc, i, no_of_bits =5 ;int ben_ind, ben_count;bool has_ben = false, mbm_test = true, mba_test = true, cqm_test = true; +bool cat_test = true;char *benchmark_cmd[BENCHMARK_ARGS ];char bw_report[64 ], bm_type[64 ];@@ -62,6 +64,7 @@ int main(int argc, char **argv) mbm_test = false; mba_test = false; cqm_test = false; + cat_test = false;while (token) {if (!strcmp (token,"mbm" )) { mbm_test = true;@@ -69,6 +72,8 @@ int main(int argc, char **argv) mba_test = true; }else if (!strcmp (token,"cqm" )) { cqm_test = true; + }else if (!strcmp (token,"cat" )) { + cat_test = true; }else {printf ("invalid argument\n" );@@ -158,6 +163,13 @@ int main(int argc, char **argv)printf ("Error in CQM test!\n" );cqm_test_cleanup (); } +if (cat_test) { +printf ("\nCAT Test Starting..\n" ); + res =cat_perf_miss_val (cpu_no, no_of_bits,"L3" ); +if (res) +printf ("Error in CAT test!\n" ); +cat_test_cleanup (); + }return 0 ; }tools/testing/selftests/resctrl/resctrlfs.c +7
-3@@ -14,12 +14,14 @@#define RESCTRL_MBM "L3 monitoring detected"#define RESCTRL_MBA "MB allocation detected"#define RESCTRL_CQM "L3 monitoring detected" +#define RESCTRL_L3_CAT "L3 allocation detected"#define CORE_SIBLINGS_PATH "/sys/bus/cpu/devices/cpu"enum {RESCTRL_FEATURE_MBM ,RESCTRL_FEATURE_MBA ,RESCTRL_FEATURE_CQM , +RESCTRL_FEATURE_L3_CAT ,MAX_RESCTRL_FEATURES };@@ -490,6 +492,7 @@ int write_schemata(char *ctrlgrp, char *schemata, int cpu_no, char *resctrl_val)FILE *fp;if ((strcmp (resctrl_val,"mba" ) ==0 ) || + (strcmp (resctrl_val,"cat" ) ==0 ) || (strcmp (resctrl_val,"cqm" ) ==0 )) {if (!schemata) {printf ("Schemata empty, so not updating\n" );@@ -507,7 +510,7 @@ int write_schemata(char *ctrlgrp, char *schemata, int cpu_no, char *resctrl_val)else sprintf (controlgroup,"%s/schemata" ,RESCTRL_PATH ); -if (!strcmp (resctrl_val,"cqm" )) +if (!strcmp (resctrl_val,"cat" ) || !strcmp (resctrl_val,"cqm" ))sprintf (schema,"%s%d%c%s" ,"L3:" , resource_id,'=' , schemata);if (strcmp (resctrl_val,"mba" ) ==0 )@@ -545,7 +548,7 @@ int validate_resctrl_feature_request(char *resctrl_val) {int resctrl_features_supported[MAX_RESCTRL_FEATURES ];const char *resctrl_features_list[MAX_RESCTRL_FEATURES ] = { -"mbm" ,"mba" ,"cqm" }; +"mbm" ,"mba" ,"cat" ,"cqm" };int i, valid_resctrl_feature =-1 ;char line[1024 ];FILE *fp;@@ -591,7 +594,8 @@ int validate_resctrl_feature_request(char *resctrl_val) resctrl_features_supported[RESCTRL_FEATURE_MBA ] =1 ;if ((strstr (line,RESCTRL_CQM )) !=NULL ) resctrl_features_supported[RESCTRL_FEATURE_CQM ] =1 ; - +if ((strstr (line,RESCTRL_L3_CAT )) !=NULL ) + resctrl_features_supported[RESCTRL_FEATURE_L3_CAT ] =1 ; }fclose (fp);
PATCH v7 10/13 selftests/resctrl: Add vendor detection mechanism
In reply to Fenghua YuFrom: Babu Moger <[email protected]>
RESCTRL feature is supported both on Intel and AMD now. Some features are implemented differently. Add vendor detection mechanism. Use the vendor check where there are differences.
--- tools/testing/selftests/resctrl/resctrl.h | 13 +++++ tools/testing/selftests/resctrl/resctrl_tests.c | 63 +++++++++++++++++++++++++ 2 files changed, 76 insertions(+)
tools/testing/selftests/resctrl/resctrl.h +13 -0
@@ -63,10 +63,23 @@ struct resctrl_val_param { }; +/** +* Results of CPUID operation are stored in this structure. +* It consists of 4x32bits IA registers: EAX, EBX, ECX and EDX. +*/ +struct cpuid_out { +uint32_t eax; +uint32_t ebx; +uint32_t ecx; +uint32_t edx; +}; +pid_t bm_pid, ppid;extern char cbm_mask[256 ];extern unsigned long long_mask;extern char llc_occup_path[1024 ]; +extern int genuine_intel; +extern int authentic_amd;int remount_resctrlfs (bool mum_resctrlfs);int get_resource_id (int cpu_no,int *resource_id);
tools/testing/selftests/resctrl/resctrl_tests.c +63 -0
75 lines not on this page · raw message
Signature · 1 line
2.7.4
Re: [PATCH v7 10/13] selftests/resctrl: Add vendor detection mechanism
On Sat, 9 Feb 2019 18:50:39 -0800
Fenghua Yu <[email protected]> wrote:
Hi,
Quoted 5 lines
From: Babu Moger <[email protected]>
RESCTRL feature is supported both on Intel and AMD now. Some features are implemented differently. Add vendor detection mechanism. Use the vendor check where there are differences.
I don't think vendor detection is the right approach. The Linux userland interface should be even architecture agnostic, not to speak of different vendors.
But even if we need this for some reason ...
Quoted 23 lines
Signed-off-by Babu Moger <[email protected]>Signed-off-by Fenghua Yu <[email protected]>--- tools/testing/selftests/resctrl/resctrl.h | 13 +++++ tools/testing/selftests/resctrl/resctrl_tests.c | 63 +++++++++++++++++++++++++ 2 files changed, 76 insertions(+)tools/testing/selftests/resctrl/resctrl.h +7
-0@@ -63,10 +63,23 @@ struct resctrl_val_param { }; +/** +* Results of CPUID operation are stored in this structure. +* It consists of 4x32bits IA registers: EAX, EBX, ECX and EDX. +*/ +struct cpuid_out { +uint32_t eax; +uint32_t ebx;
tools/testing/selftests/resctrl/resctrl.h struct resctrl_val_param { · line 82
Quoted 10 lines
tools/testing/selftests/resctrl/resctrl.h +6
-0+uint32_t ecx; +uint32_t edx; +}; +pid_t bm_pid, ppid;extern char cbm_mask[256 ];extern unsigned long long_mask;extern char llc_occup_path[1024 ]; +extern int genuine_intel; +extern int authentic_amd;
There are more CPU vendors than that ;-)
And to not encourage this kind of vendor specific tests, I'd keep this exposure to a minimum.
In my version I have a simple is_amd() function, to cover the L3 vs. physical pacakges differences.
Quoted 9 lines
tools/testing/selftests/resctrl/resctrl.h +0
-0int remount_resctrlfs (bool mum_resctrlfs);int get_resource_id (int cpu_no,int *resource_id);tools/testing/selftests/resctrl/resctrl_tests.c +0
-0@@ -14,6 +14,66 @@#define BENCHMARK_ARGS 64
tools/testing/selftests/resctrl/resctrl_tests.c line 24
Quoted 10 lines
tools/testing/selftests/resctrl/resctrl_tests.c +8
-0#define BENCHMARK_ARG_SIZE 64 +/** +* This variables provides information about the vendor +*/ +int genuine_intel; +int authentic_amd; + +void lcpuid (const unsigned int leaf,const unsigned int subleaf, +struct cpuid_out *out)
There is a much easier way to detect the vendor, without resorting to (unchecked) inline assembly in userland:
I changed this to scan /proc/cpuinfo for a line starting with vendor_id, then use the information there. This should work everywhere.
Cheers,
Andre
Quoted 65 lines
tools/testing/selftests/resctrl/resctrl_tests.c +55
-0+{ +if (!out) +return ; + +#ifdef __x86_64__ + asmvolatile ("mov %4, %%eax\n\t" +"mov %5, %%ecx\n\t" +"cpuid\n\t" +"mov %%eax, %0\n\t" +"mov %%ebx, %1\n\t" +"mov %%ecx, %2\n\t" +"mov %%edx, %3\n\t" + :"=g" (out->eax),"=g" (out->ebx),"=g" (out->ecx), +"=g" (out->edx) + :"g" (leaf),"g" (subleaf) + :"%eax" ,"%ebx" ,"%ecx" ,"%edx" ); +#else + asmvolatile ("push %%ebx\n\t" +"mov %4, %%eax\n\t" +"mov %5, %%ecx\n\t" +"cpuid\n\t" +"mov %%eax, %0\n\t" +"mov %%ebx, %1\n\t" +"mov %%ecx, %2\n\t" +"mov %%edx, %3\n\t" +"pop %%ebx\n\t" + :"=g" (out->eax),"=g" (out->ebx),"=g" (out->ecx), +"=g" (out->edx) + :"g" (leaf),"g" (subleaf) + :"%eax" ,"%ecx" ,"%edx" ); +#endif +} + +int detect_vendor (void ) +{ +struct cpuid_out vendor; +int ret =0 ; + +lcpuid (0x0 ,0x0 , &vendor); +if (vendor.ebx ==0x756e6547 && vendor.edx ==0x49656e69 && + vendor.ecx ==0x6c65746e ) { + genuine_intel =1 ; + }else if (vendor.ebx ==0x68747541 && vendor.edx ==0x69746E65 && + vendor.ecx ==0x444D4163 ) { + authentic_amd =1 ; + }else { + ret = -EFAULT ; + } + +return ret; +} +static void cmd_help (void ) {printf ("usage: resctrl_tests [-h] [-b \"benchmark_cmd [options]\"] [-t test list] [-n no_of_bits]\n" );@@ -110,6 +170,9 @@ int main(int argc, char **argv)return errno; } +/* Detect vendor */ +detect_vendor (); +if (has_ben) {/* Extract benchmark command from command line. */ for (i = ben_ind; i < argc; i++) {
James Morse
reviewerRe: [PATCH v7 10/13] selftests/resctrl: Add vendor detection mechanism
Hi Andre,
(thanks for digging into this!)
Quoted 34 lines · Andre Przywara
Quoted 5 lines · On Sat, 9 Feb 2019 18:50:39 -0800 Fenghua Yu <[email protected]> wrote:
From: Babu Moger <[email protected]>
RESCTRL feature is supported both on Intel and AMD now. Some features are implemented differently. Add vendor detection mechanism. Use the vendor check where there are differences.
I don't think vendor detection is the right approach. The Linux userland interface should be even architecture agnostic, not to speak of different vendors.
But even if we need this for some reason ...
Quoted 6 lines
tools/testing/selftests/resctrl/resctrl_tests.c +0
-0@@ -14,6 +14,66 @@#define BENCHMARK_ARGS 64tools/testing/selftests/resctrl/resctrl_tests.c line 24
Quoted 10 lines
tools/testing/selftests/resctrl/resctrl_tests.c +8
-0#define BENCHMARK_ARG_SIZE 64 +/** +* This variables provides information about the vendor +*/ +int genuine_intel; +int authentic_amd; + +void lcpuid (const unsigned int leaf,const unsigned int subleaf, +struct cpuid_out *out)There is a much easier way to detect the vendor, without resorting to (unchecked) inline assembly in userland:
Quoted 2 lines
I changed this to scan
/proc/cpuinfofor a line starting withvendor_id, then use the information there. This should work everywhere.
Everywhere x86. /proc/cpuinfo is unfortunately arch specific, arm64 spells that field 'CPU implementer'. Short of invoking lscpu, I don't know a portable way of finding this string.
What do we need it for? Surely it indicates something is wrong with the kernel interface if you need to know which flavour of CPU this is.
The only differences I've spotted are the 'MAX_MBA_BW' on Intel is a percentage, on AMD its a number between 1 and 2K. If user-space needs to know we could add another file with the 'max_bandwidth'. (I haven't spotted how the MBA default_ctrl gets exposed).
The other one is the non-contiguous CBM values, where user-space is expected to try it and see [0]. If user-space needs to know, we could expose some 'sparse_bitmaps=[0 1]', unaware user-space keeps working.
Thanks,
James
Re: [PATCH v7 10/13] selftests/resctrl: Add vendor detection mechanism
On 14/05/2019 18:20, James Morse wrote:
Hi James,
Quoted 37 lines
(thanks for digging into this!)
Quoted 33 lines · Andre Przywara
Quoted 5 lines · On Sat, 9 Feb 2019 18:50:39 -0800 Fenghua Yu <[email protected]> wrote:
From: Babu Moger <[email protected]>
RESCTRL feature is supported both on Intel and AMD now. Some features are implemented differently. Add vendor detection mechanism. Use the vendor check where there are differences.
I don't think vendor detection is the right approach. The Linux userland interface should be even architecture agnostic, not to speak of different vendors.
But even if we need this for some reason ...
Quoted 6 lines
tools/testing/selftests/resctrl/resctrl_tests.c +0
-0@@ -14,6 +14,66 @@#define BENCHMARK_ARGS 64tools/testing/selftests/resctrl/resctrl_tests.c line 24
Quoted 10 lines
tools/testing/selftests/resctrl/resctrl_tests.c +8
-0#define BENCHMARK_ARG_SIZE 64 +/** +* This variables provides information about the vendor +*/ +int genuine_intel; +int authentic_amd; + +void lcpuid (const unsigned int leaf,const unsigned int subleaf, +struct cpuid_out *out)There is a much easier way to detect the vendor, without resorting to (unchecked) inline assembly in userland:
Quoted 5 lines
Quoted 2 lines
I changed this to scan
/proc/cpuinfofor a line starting withvendor_id, then use the information there. This should work everywhere.Everywhere x86.
/proc/cpuinfois unfortunately arch specific, arm64 spells that field 'CPU implementer'. Short of invoking lscpu, I don't know a portable way of finding this string.
Well, what I meant is: It works everywhere to tell whether this box has an x86 AMD CPU. Yes, it probably won't match anything on other architectures, but that's fine, as at least it will compile and won't crash.
Quoted 2 lines
What do we need it for? Surely it indicates something is wrong with the kernel interface if you need to know which flavour of CPU this is.
As you mentioned, we should not need it. I just couldn't find a better way (yet) to differentiate between L3 cache ID and physical package ID (see patch 11/13). So this is a kludge for now to not break this particular code.
Out of curiosity: Is there any userland tool meant to control the resources? I guess poking around in sysfs is not how admins are expected to use this?
This tool would surely run into the same problems, which somewhat tell me that the interface is not really right.
Cheers,
Andre.
Quoted 15 lines
The only differences I've spotted are the '
MAX_MBA_BW' on Intel is a percentage, on AMD its a number between 1 and 2K. If user-space needs to know we could add another file with the 'max_bandwidth'. (I haven't spotted how the MBAdefault_ctrlgets exposed).The other one is the non-contiguous CBM values, where user-space is expected to try it and see [0]. If user-space needs to know, we could expose some 'sparse_bitmaps=[0 1]', unaware user-space keeps working.
Thanks,
James
James Morse
reviewerRe: [PATCH v7 10/13] selftests/resctrl: Add vendor detection mechanism
Hi André,
Quoted 16 lines · André Przywara
Quoted 14 lines · James Morse
Quoted 13 lines · Andre Przywara
Quoted 5 lines · On Sat, 9 Feb 2019 18:50:39 -0800 Fenghua Yu <[email protected]> wrote:
From: Babu Moger <[email protected]>
RESCTRL feature is supported both on Intel and AMD now. Some features are implemented differently. Add vendor detection mechanism. Use the vendor check where there are differences.
I don't think vendor detection is the right approach. The Linux userland interface should be even architecture agnostic, not to speak of different vendors.
But even if we need this for some reason ...
Quoted 7 lines
Quoted 2 lines
What do we need it for? Surely it indicates something is wrong with the kernel interface if you need to know which flavour of CPU this is.
As you mentioned, we should not need it. I just couldn't find a better way (yet) to differentiate between L3 cache ID and physical package ID (see patch 11/13). So this is a kludge for now to not break this particular code.
[0]? That's broken. It needs to take the 'cache/index?/id' field, and not hard-code '3', search each 'cache/index?/level' instead.
Documentation/x86/resctrl_ui.rst's "Cache IDs" section says:
| On current generation systems there is one L3 cache per socket and L2 | caches are generally just shared by the hyperthreads on a core, but this | isn't an architectural requirement.
[...]
| So instead of using "socket" or "core" to define the set of logical cpus | sharing a resource we use a "Cache ID"
[...]
| To find the ID for each logical CPU look in
| /sys/devices/system/cpu/cpu*/cache/index*/id
arch/x86/kernel/cpu/restrl/core.c:domain_add_cpu() pulls the domain-id out of struct cacheinfo:
| int id = get_cache_id(cpu, r->cache_level);
drivers/base/cacheinfo.c has some macro-foliage that exports this same field via sysfs, and arch/x86/kernel/cpu/restrl/ctrlmondata.c:parse_line() matches that id against the value user-space provides in the schemata.
(we've got some horrible code for arm64 to make this work without 'cache id' as a hardware property!)
On x86 these numbers are of the order 0,1,2, so its very likely physical_package_id and cache_id alias, and you get away with it.
Quoted 3 lines
Out of curiosity: Is there any userland tool meant to control the resources? I guess poking around in sysfs is not how admins are expected to use this?
I've come across:
https://github.com/intel/intel-cmt-cat/
but I've never even cloned it. The rdtset man page has:
| With --iface-os (-I) parameter, rdtset uses resctrl filesystem (/sys/fs/resctrl) | instead of accessing MSRs directly.
Quoted 2 lines
This tool would surely run into the same problems, which somewhat tell me that the interface is not really right.
At the moment its not as-documented or as the kernel is using those numbers.
I assume this is something that changed in resctrl when it was merged, and this selftest tool just needs updating.
Thanks,
James
PATCH v7 11/13 selftests/resctrl: Use cache index3 id for AMD schemata masks
In reply to Fenghua YuFrom: Babu Moger <[email protected]>
AMD uses the cache l3 boundary for schemata masks. Update it accordigly.
--- tools/testing/selftests/resctrl/resctrlfs.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)
tools/testing/selftests/resctrl/resctrlfs.c +7 -2
@@ -110,8 +110,13 @@ int get_resource_id(int cpu_no, int *resource_id)char phys_pkg_path[1024 ];FILE *fp; -sprintf (phys_pkg_path,"%s%d/topology/physical_package_id" , -PHYS_ID_PATH , cpu_no); +if (authentic_amd) +sprintf (phys_pkg_path,"%s%d/cache/index3/id" , +PHYS_ID_PATH , cpu_no); +else +sprintf (phys_pkg_path,"%s%d/topology/physical_package_id" , +PHYS_ID_PATH , cpu_no); + fp =fopen (phys_pkg_path,"r" );if (!fp) {perror ("Failed to open physical_package_id" );
Signature · 1 line
2.7.4
PATCH v7 12/13 selftests/resctrl: Disable MBA and MBM tests for AMD
In reply to Fenghua YuFrom: Babu Moger <[email protected]>
For now, disable MBA and MBM tests for AMD. Deciding test pass/fail is not clear right now. We can enable when we have some clarity.
--- tools/testing/selftests/resctrl/cat_test.c | 2 +- tools/testing/selftests/resctrl/resctrl_tests.c | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-)
tools/testing/selftests/resctrl/cat_test.c +1 -1
@@ -69,7 +69,7 @@ static void show_cache_info(unsigned long sum_llc_perf_miss, int no_of_bits,printf ("Allocated cache lines: %lu \t" , allocated_cache_lines);printf ("Percent diff=%d \t" ,abs ((int )diff_percent)); -if (abs ((int )diff_percent) >MAX_DIFF_PERCENT ) +if (genuine_intel && (abs ((int )diff_percent) >MAX_DIFF_PERCENT ))printf ("Failed\n" );else printf ("Passed\n" );
tools/testing/selftests/resctrl/resctrl_tests.c +2 -2
@@ -197,7 +197,7 @@ int main(int argc, char **argv)sprintf (bw_report,"reads" );sprintf (bm_type,"fill_buf" ); -if (mbm_test) { +if (genuine_intel && mbm_test) {printf ("\nMBM BW Change Starting..\n" );if (!has_ben)sprintf (benchmark_cmd[5 ],"%s" ,"mbm" );
@@ -207,7 +207,7 @@ int main(int argc, char **argv)mbm_test_cleanup (); } -if (mba_test) { +if (genuine_intel && mba_test) {printf ("\nMBA Schemata Change Starting..\n" );if (!has_ben)sprintf (benchmark_cmd[5 ],"%s" ,"mba" );
Signature · 1 line
2.7.4
Re: [PATCH v7 12/13] selftests/resctrl: Disable MBA and MBM tests for AMD
On Sat, 9 Feb 2019 18:50:41 -0800
Fenghua Yu <[email protected]> wrote:
Hi,
Quoted 4 lines
From: Babu Moger <[email protected]>
For now, disable MBA and MBM tests for AMD. Deciding test pass/fail is not clear right now. We can enable when we have some clarity.
I don't think this is the right way. The availability of features should be queryable, for instance by looking into /sys/fs/resctrl/info. Checking for a certain vendor to skip tests just sounds wrong to me, and is definitely not scalable or future proof.
We should really check the availability of a feature, then skip the whole subsystem test in resctrl_tests.c.
Cheers,
Andre.
Quoted 43 lines
Signed-off-by Babu Moger <[email protected]>Signed-off-by Fenghua Yu <[email protected]>--- tools/testing/selftests/resctrl/cat_test.c | 2 +- tools/testing/selftests/resctrl/resctrl_tests.c | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-)tools/testing/selftests/resctrl/cat_test.c +1
-1@@ -69,7 +69,7 @@ static void show_cache_info(unsigned long sum_llc_perf_miss, int no_of_bits,printf ("Allocated cache lines: %lu \t" , allocated_cache_lines);printf ("Percent diff=%d \t" ,abs ((int )diff_percent)); -if (abs ((int )diff_percent) >MAX_DIFF_PERCENT ) +if (genuine_intel && (abs ((int )diff_percent) >MAX_DIFF_PERCENT ))printf ("Failed\n" );else printf ("Passed\n" );tools/testing/selftests/resctrl/resctrl_tests.c +2
-2@@ -197,7 +197,7 @@ int main(int argc, char **argv)sprintf (bw_report,"reads" );sprintf (bm_type,"fill_buf" ); -if (mbm_test) { +if (genuine_intel && mbm_test) {printf ("\nMBM BW Change Starting..\n" );if (!has_ben)sprintf (benchmark_cmd[5 ],"%s" ,"mbm" );@@ -207,7 +207,7 @@ int main(int argc, char **argv)mbm_test_cleanup (); } -if (mba_test) { +if (genuine_intel && mba_test) {printf ("\nMBA Schemata Change Starting..\n" );if (!has_ben)sprintf (benchmark_cmd[5 ],"%s" ,"mba" );
RE: [PATCH v7 12/13] selftests/resctrl: Disable MBA and MBM tests for AMD
Quoted 25 lines
On Friday, May 10, 2019 10:40 AM
Andre Przywara [mailto:[email protected]] wrote:
To: Yu, Fenghua <[email protected]>
Cc: Thomas Gleixner <[email protected]>; Ingo Molnar <[email protected]>; Borislav Petkov <[email protected]>; H Peter Anvin <[email protected]>; Luck, Tony <[email protected]>; Chatre, Reinette <[email protected]>; Shankar, Ravi V <[email protected]>; Shen, Xiaochen <[email protected]>; Pathan, Arshiya Hayatkhan <[email protected]>; Prakhya, Sai Praneeth <[email protected]>; Babu Moger <[email protected]>; linux-kernel <[email protected]>; James Morse <[email protected]>
Subject: Re: [PATCH v7 12/13] selftests/resctrl: Disable MBA and MBM tests for AMDOn Sat, 9 Feb 2019 18:50:41 -0800
Fenghua Yu <[email protected]> wrote:Hi,
Quoted 4 lines
From: Babu Moger <[email protected]>
For now, disable MBA and MBM tests for AMD. Deciding test pass/fail is not clear right now. We can enable when we have some clarity.
Quoted 7 lines
I don't think this is the right way. The availability of features should be queryable, for instance by looking into
/sys/fs/resctrl/info. Checking for a certain vendor to skip tests just sounds wrong to me, and is definitely not scalable or future proof.We should really check the availability of a feature, then skip the whole subsystem test in
resctrl_tests.c.
Babu may correct if I'm wrong: AMD does support the MBA and MBM features. So if querying the info directory, the features are there. But AMD doesn't want to support the testing for the features in the current patch set. They may support the testing in the future.
Thanks.
-Fenghua
Moger, Babu
reviewerRe: [PATCH v7 12/13] selftests/resctrl: Disable MBA and MBM tests for AMD
Fenghua/Andre,
Quoted 32 lines · Fenghua
[CAUTION: External Email]
Quoted 30 lines
On Friday, May 10, 2019 10:40 AM
Andre Przywara [mailto:[email protected]] wrote:
To: Yu, Fenghua <[email protected]>
Cc: Thomas Gleixner <[email protected]>; Ingo Molnar <[email protected]>; Borislav Petkov <[email protected]>; H Peter Anvin <[email protected]>; Luck, Tony <[email protected]>; Chatre, Reinette <[email protected]>; Shankar, Ravi V <[email protected]>; Shen, Xiaochen <[email protected]>; Pathan, Arshiya Hayatkhan <[email protected]>; Prakhya, Sai Praneeth <[email protected]>; Babu Moger <[email protected]>; linux-kernel <[email protected]>; James Morse <[email protected]>
Subject: Re: [PATCH v7 12/13] selftests/resctrl: Disable MBA and MBM tests for AMDOn Sat, 9 Feb 2019 18:50:41 -0800
Fenghua Yu <[email protected]> wrote:Hi,
Quoted 4 lines
From: Babu Moger <[email protected]>
For now, disable MBA and MBM tests for AMD. Deciding test pass/fail is not clear right now. We can enable when we have some clarity.
I don't think this is the right way. The availability of features should be queryable, for instance by looking into
/sys/fs/resctrl/info. Checking for a certain vendor to skip tests just sounds wrong to me, and is definitely not scalable or future proof.
Quoted 4 lines
Quoted 2 lines
We should really check the availability of a feature, then skip the whole subsystem test in
resctrl_tests.c.Babu may correct if I'm wrong: AMD does support the MBA and MBM features. So if querying the info directory, the features are there. But AMD doesn't want to support the testing for the features in the current patch set. They may support the testing in the future.
Yes. AMD supports the MBA and MBM features. But, deciding the test pass/fail was not clear to me. That is why I decided to disable the test for now. Right now, it reports the test result as "fail". We can add the test in the future with some modifications.
Quoted 5 lines
Thanks.
-Fenghua
PATCH v7 13/13 selftests/resctrl: Add the test in MAINTAINERS
In reply to Fenghua YuThe resctrl selftest will be maintained by RDT maintainers.
--- MAINTAINERS | 1 + 1 file changed, 1 insertion(+)
MAINTAINERS +1 -0
@@ -12826,6 +12826,7 @@ S: Supported F: arch/x86/kernel/cpu/resctrl/ F: arch/x86/include/asm/resctrl_sched.h F: Documentation/x86/resctrl* +F: tools/testing/selftests/resctrl/ READ-COPY UPDATE (RCU) M: "Paul E. McKenney" <[email protected]>
Signature · 1 line
2.7.4
Quoted 3 lines · Fenghua Yu
With more and more resctrl features are being added by Intel, AMD and ARM, a test tool is becoming more and more useful to validate that both hardware and software functionalities work as expected.
Hi, Thomas, Ingo, and Borislav,
Do you have any comments on the resctrl selftest tool?
Do you have any plan to push the patches to upstream?
The tool has been tested on both Intel RDT and AMD QoS.
Thanks.
-Fenghua
Moger, Babu
reviewerHi Maintainers,
Ping again. Any concerns with these patches?
Thanks
Babu
Quoted 29 lines
-----Original Message-----
From: Fenghua Yu <[email protected]>
Sent: Thursday, February 14, 2019 8:56 AM
To: Thomas Gleixner <[email protected]>; Ingo Molnar <[email protected]>; Borislav Petkov <[email protected]>; H Peter Anvin <[email protected]>; Tony Luck <[email protected]>; Reinette Chatre <[email protected]>; Ravi V Shankar <[email protected]>; Xiaochen Shen <[email protected]>; Arshiya Hayatkhan Pathan <[email protected]>; Sai Praneeth Prakhya <[email protected]>; Moger, Babu <[email protected]>
Cc: linux-kernel <[email protected]>
Subject: Re: [PATCH v7 00/13] selftests/resctrl: Add resctrl selftestQuoted 3 lines · Fenghua Yu
With more and more resctrl features are being added by Intel, AMD and ARM, a test tool is becoming more and more useful to validate that both hardware and software functionalities work as expected.
Hi, Thomas, Ingo, and Borislav,
Do you have any comments on the resctrl selftest tool?
Do you have any plan to push the patches to upstream?
The tool has been tested on both Intel RDT and AMD QoS.
Thanks.
-Fenghua
Moger, Babu
reviewerPatches look good. Tested v7 again on AMD system. Thanks
Quoted 135 lines
-----Original Message-----
From: Fenghua Yu <[email protected]>
Sent: Saturday, February 9, 2019 8:50 PM
To: Thomas Gleixner <[email protected]>; Ingo Molnar <[email protected]>; Borislav Petkov <[email protected]>; H Peter Anvin <[email protected]>; Tony Luck <[email protected]>; Reinette Chatre <[email protected]>; Ravi V Shankar <[email protected]>; Xiaochen Shen <[email protected]>; Arshiya Hayatkhan Pathan <[email protected]>; Sai Praneeth Prakhya <[email protected]>; Moger, Babu <[email protected]>
Cc: linux-kernel <[email protected]>; Fenghua Yu <[email protected]>
Subject: [PATCH v7 00/13] selftests/resctrl: Add resctrl selftestWith more and more resctrl features are being added by Intel, AMD and ARM, a test tool is becoming more and more useful to validate that both hardware and software functionalities work as expected.
We introduce resctrl selftest to cover resctrl features on both X86 and ARM architectures. It first implements MBM (Memory Bandwidth Monitoring), MBA (Memory Bandwidth Allocation), L3 CAT (Cache Allocation Technology), and CQM (Cache QoS Monitoring) tests. We can enhance the selftest tool to include more functionality tests in future.
The tool has been tested on both Intel RDT and AMD QoS.
There is an existing resctrl test suite '
intel_cmt_cat'. But the major purpose of the tool is to test Intel(R) RDT hardware via writing and reading MSR registers. It does access resctrl file system; but the functionalities are very limited. And it doesn't support automatic test and a lot of manual verifications are involved.So the selftest tool we are introducing here provides a convenient tool which does automatic resctrl testing, is easily available in kernel tree, and will be extended to AMD QoS and ARM MPAM.
The selftest tool is in
tools/testing/selftests/resctrlin order to have generic test code for all architectures.Changelog:
v7:
- Fix a few warnings when compiling patches separately, pointed by Babuv6:
- Fix a benchmark reading optimized out issue in newer GCC.
- Fix a few coding style issues.
- Re-arrange code among patches to make cleaner code. No change in patches
structure.v5:
- Based the v4 patches submitted by Fenghua Yu and added changes to support
AMD.
- Changed the function nameget_sock_numtoget_resource_id. Intel uses socket number for schemata and AMD uses l3 index id. To generalize, changed the function name toget_resource_id.
- Added the code to detect vendor.
- Disabled the few tests for AMD where the test results are not clear. Also AMD does not have IMC.
- Fixed few compile issues.
- Some cleanup to make each patch independent.
- Tested the patches on AMD system. Fenghua, Need your help to test on Intel box. Please feel free to change and resubmit if something broken.
- Here is the link for previous version.
https://lore.kernel.org/lkml/1545438038-75107-1-git-send-email-
fenghua.yu=
@intel.com/v4:
- address comments from Balu and Randy
- Add CAT and CQM testsv3:
- Change code based on comments from Babu Moger
- Remove some unnessary code and use pipe to communicate b/w processesv2:
- Change code based on comments from Babu Moger
- Clean up other places.Arshiya Hayatkhan Pathan (4):
selftests/resctrl: Add MBM test
selftests/resctrl: Add MBA test
selftests/resctrl: Add Cache QoS Monitoring (CQM) selftest selftests/resctrl: Add Cache Allocation Technology (CAT) selftestBabu Moger (3):
selftests/resctrl: Add vendor detection mechanism selftests/resctrl: Use cache index3 id for AMD schemata masks selftests/resctrl: Disable MBA and MBM tests for AMDFenghua Yu (2):
selftests/resctrl: Add README for resctrl tests
selftests/resctrl: Add the test inMAINTAINERSSai Praneeth Prakhya (4):
selftests/resctrl: Add basic resctrl file system operations and data selftests/resctrl: Read memory bandwidth from perf IMC counter and
from resctrl file system
selftests/resctrl: Add callback to start a benchmark selftests/resctrl: Add built in benchmarkMAINTAINERS | 1 + tools/testing/selftests/resctrl/Makefile | 16 + tools/testing/selftests/resctrl/README | 54 ++ tools/testing/selftests/resctrl/cache.c | 274 +++++++++ tools/testing/selftests/resctrl/cat_test.c | 242 ++++++++ tools/testing/selftests/resctrl/cqm_test.c | 173 ++++++ tools/testing/selftests/resctrl/fill_buf.c | 210 +++++++ tools/testing/selftests/resctrl/mba_test.c | 178 ++++++ tools/testing/selftests/resctrl/mbm_test.c | 150 +++++ tools/testing/selftests/resctrl/resctrl.h | 117 ++++ tools/testing/selftests/resctrl/resctrl_tests.c | 238 ++++++++ tools/testing/selftests/resctrl/resctrl_val.c | 723 ++++++++++++++++++++++++ tools/testing/selftests/resctrl/resctrlfs.c | 668 ++++++++++++++++++++++ 13 files changed, 3044 insertions(+) create mode 100644 tools/testing/selftests/resctrl/Makefile create mode 100644 tools/testing/selftests/resctrl/README create mode 100644 tools/testing/selftests/resctrl/cache.c create mode 100644 tools/testing/selftests/resctrl/cat_test.c create mode 100644 tools/testing/selftests/resctrl/cqm_test.c create mode 100644 tools/testing/selftests/resctrl/fill_buf.c create mode 100644 tools/testing/selftests/resctrl/mba_test.c create mode 100644 tools/testing/selftests/resctrl/mbm_test.c create mode 100644 tools/testing/selftests/resctrl/resctrl.h create mode 100644 tools/testing/selftests/resctrl/resctrl_tests.c create mode 100644 tools/testing/selftests/resctrl/resctrl_val.c create mode 100644 tools/testing/selftests/resctrl/resctrlfs.c--
2.7.4
On Sat, 9 Feb 2019 18:50:29 -0800
Fenghua Yu <[email protected]> wrote:
Hi Fenghua, Babu,
Quoted 3 lines
With more and more resctrl features are being added by Intel, AMD and ARM, a test tool is becoming more and more useful to validate that both hardware and software functionalities work as expected.
That's very much appreciated! We decided to use that tool here to detect regressions in James' upcoming resctrl rework series. While doing so we spotted some shortcomings:
- There is some unconditional x86 inline assembly which obviously breaks the build on ARM.
- The result output is somewhat confusing, I find it hard to see whether a test succeeded or not, also it's not easily parseable by scripts.
- There are some basic tests missing (does the filesystem exist? Is the mount point visible?)
- Some tests create files in the current directory. This is somewhat confusing if you happen to be in some sysfs directory, for instance. I don't think we really need temporary files (pipes should cover most of the cases), but in case we do, we should create them in some /tmp directory, probably by using a glibc function for this.
- There were some problems that newer GCC and clang complained about.
For now I didn't look too closely at the actual test algorithms, but decided to fix those things first. I replied to some of the smaller issues I spotted on the way in the individual patch mails.
For the bigger things (new tests, result formatting) I just went ahead and fixed them in my own branch [1] (work in progress!). I went with TAP formatting, because this is both readable by humans and can be processed by the "prove" tool (or any other simple UNIX tool, for that matter).
I used this amended version to do some regression testing on James' resctrl rework series, which involved boot testing more than 100 patches, so automation is key here. The result looks like this:
===================
TAP version 13
ok kernel supports resctrl filesystem
ok resctrl mountpoint "/sys/fs/resctrl" exists
# resctrl filesystem not mounted # Starting MBM BW change ... ok MBM: bw change # SKIP memory bandwidth not supported # Starting MBA Schemata Change ... ok MBA: change-schemata # SKIP memory bandwidth not supported # Starting CQM test ... ok mounting resctrl to "/sys/fs/resctrl" ok CQM: test # SKIP L3 not supported # Starting CAT test ... ok mounting resctrl to "/sys/fs/resctrl" cache size :31457280 ok writing benchmark parameters to resctrl FS ok Write schema "L3:0=7fff" to resctrl FS ok cache difference more than 4 % # Percent diff=16 # Number of bits: 15 # Avg_llc_perf_miss: 306994 # Allocated cache lines: 368640 ok CAT: test 1..11 =================
Appreciate any feedback on this!
Cheers,
Andre.
[1] http://linux-arm.org/git?p=linux-ap.git;a=shortlog;h=refs/heads/resctrl-selftests-wipgit://linux-arm.org/linux-ap.git branch: resctrl-selftests-wip
Quoted 110 lines
We introduce resctrl selftest to cover resctrl features on both X86 and ARM architectures. It first implements MBM (Memory Bandwidth Monitoring), MBA (Memory Bandwidth Allocation), L3 CAT (Cache Allocation Technology), and CQM (Cache QoS Monitoring) tests. We can enhance the selftest tool to include more functionality tests in future.
The tool has been tested on both Intel RDT and AMD QoS.
There is an existing resctrl test suite '
intel_cmt_cat'. But the major purpose of the tool is to test Intel(R) RDT hardware via writing and reading MSR registers. It does access resctrl file system; but the functionalities are very limited. And it doesn't support automatic test and a lot of manual verifications are involved.So the selftest tool we are introducing here provides a convenient tool which does automatic resctrl testing, is easily available in kernel tree, and will be extended to AMD QoS and ARM MPAM.
The selftest tool is in
tools/testing/selftests/resctrlin order to have generic test code for all architectures.Changelog:
v7:
- Fix a few warnings when compiling patches separately, pointed by Babuv6:
- Fix a benchmark reading optimized out issue in newer GCC.
- Fix a few coding style issues.
- Re-arrange code among patches to make cleaner code. No change in patches structure.v5:
- Based the v4 patches submitted by Fenghua Yu and added changes to support AMD.
- Changed the function nameget_sock_numtoget_resource_id. Intel uses socket number for schemata and AMD uses l3 index id. To generalize, changed the function name toget_resource_id.
- Added the code to detect vendor.
- Disabled the few tests for AMD where the test results are not clear. Also AMD does not have IMC.
- Fixed few compile issues.
- Some cleanup to make each patch independent.
- Tested the patches on AMD system. Fenghua, Need your help to test on Intel box. Please feel free to change and resubmit if something broken.
- Here is the link for previous version.
https://lore.kernel.org/lkml/1545438038-75107-1-git-send-email-fenghua.yu=
@intel.com/v4:
- address comments from Balu and Randy
- Add CAT and CQM testsv3:
- Change code based on comments from Babu Moger
- Remove some unnessary code and use pipe to communicate b/w processesv2:
- Change code based on comments from Babu Moger
- Clean up other places.Arshiya Hayatkhan Pathan (4):
selftests/resctrl: Add MBM test
selftests/resctrl: Add MBA test
selftests/resctrl: Add Cache QoS Monitoring (CQM) selftest selftests/resctrl: Add Cache Allocation Technology (CAT) selftestBabu Moger (3):
selftests/resctrl: Add vendor detection mechanism selftests/resctrl: Use cache index3 id for AMD schemata masks selftests/resctrl: Disable MBA and MBM tests for AMDFenghua Yu (2):
selftests/resctrl: Add README for resctrl tests
selftests/resctrl: Add the test inMAINTAINERSSai Praneeth Prakhya (4):
selftests/resctrl: Add basic resctrl file system operations and data selftests/resctrl: Read memory bandwidth from perf IMC counter and
from resctrl file system
selftests/resctrl: Add callback to start a benchmark selftests/resctrl: Add built in benchmarkMAINTAINERS | 1 + tools/testing/selftests/resctrl/Makefile | 16 + tools/testing/selftests/resctrl/README | 54 ++ tools/testing/selftests/resctrl/cache.c | 274 +++++++++ tools/testing/selftests/resctrl/cat_test.c | 242 ++++++++ tools/testing/selftests/resctrl/cqm_test.c | 173 ++++++ tools/testing/selftests/resctrl/fill_buf.c | 210 +++++++ tools/testing/selftests/resctrl/mba_test.c | 178 ++++++ tools/testing/selftests/resctrl/mbm_test.c | 150 +++++ tools/testing/selftests/resctrl/resctrl.h | 117 ++++ tools/testing/selftests/resctrl/resctrl_tests.c | 238 ++++++++ tools/testing/selftests/resctrl/resctrl_val.c | 723 ++++++++++++++++++++++++ tools/testing/selftests/resctrl/resctrlfs.c | 668 ++++++++++++++++++++++ 13 files changed, 3044 insertions(+) create mode 100644 tools/testing/selftests/resctrl/Makefile create mode 100644 tools/testing/selftests/resctrl/README create mode 100644 tools/testing/selftests/resctrl/cache.c create mode 100644 tools/testing/selftests/resctrl/cat_test.c create mode 100644 tools/testing/selftests/resctrl/cqm_test.c create mode 100644 tools/testing/selftests/resctrl/fill_buf.c create mode 100644 tools/testing/selftests/resctrl/mba_test.c create mode 100644 tools/testing/selftests/resctrl/mbm_test.c create mode 100644 tools/testing/selftests/resctrl/resctrl.h create mode 100644 tools/testing/selftests/resctrl/resctrl_tests.c create mode 100644 tools/testing/selftests/resctrl/resctrl_val.c create mode 100644 tools/testing/selftests/resctrl/resctrlfs.c
Quoted 7 lines
On Friday, May 10, 2019 10:36 AM
Andre Przywara [mailto:[email protected]] wrote:
On Sat, 9 Feb 2019 18:50:29 -0800
Fenghua Yu <[email protected]> wrote:Hi Fenghua, Babu,
Quoted 7 lines
Quoted 3 lines
With more and more resctrl features are being added by Intel, AMD and ARM, a test tool is becoming more and more useful to validate that both hardware and software functionalities work as expected.
That's very much appreciated! We decided to use that tool here to detect regressions in James' upcoming resctrl rework series. While doing so we spotted some shortcomings:
This is great!
Quoted 2 lines
- There is some unconditional x86 inline assembly which obviously breaks the build on ARM.
Will fix this as much as possible.
BTW, does ARM support perf imc_count events which are used in CAT tests?
Quoted 4 lines
- The result output is somewhat confusing, I find it hard to see whether a test succeeded or not, also it's not easily parseable by scripts.
- There are some basic tests missing (does the filesystem exist? Is the mount point visible?)
Is it OK to fold the basic tests into bigger tests? E.g. the filesystem and mount point tests are actually part of CAT test and we just dump the sub-test results during CAT test. Seems your test results show the sub-test results during CAT test.
Quoted 5 lines
- Some tests create files in the current directory. This is somewhat confusing if you happen to be in some sysfs directory, for instance. I don't think we really need temporary files (pipes should cover most of the cases), but in case we do, we should create them in some /tmp directory, probably by using a glibc function for this.
You are right. Will generate the temp file in /tmp.
Quoted 1 line
- There were some problems that newer GCC and clang complained about.
Which GCC do you use to see the issues? We use 8.2.1 and no compilation issue is reported.
Quoted 42 lines
For now I didn't look too closely at the actual test algorithms, but decided to fix those things first. I replied to some of the smaller issues I spotted on the way in the individual patch mails.
For the bigger things (new tests, result formatting) I just went ahead and fixed them in my own branch [1] (work in progress!). I went with TAP formatting, because this is both readable by humans and can be processed by the "prove" tool (or any other simple UNIX tool, for that matter).
I used this amended version to do some regression testing on James' resctrl rework series, which involved boot testing more than 100 patches, so automation is key here. The result looks like this:
===================
TAP version 13
ok kernel supports resctrl filesystem
ok resctrl mountpoint "/sys/fs/resctrl" exists # resctrl filesystem not mounted # Starting MBM BW change ...
ok MBM: bw change # SKIP memory bandwidth not supported # Starting MBA Schemata Change ...
ok MBA: change-schemata # SKIP memory bandwidth not supported # Starting CQM test ...
ok mounting resctrl to "/sys/fs/resctrl"
ok CQM: test # SKIP L3 not supported
# Starting CAT test ...
ok mounting resctrl to "/sys/fs/resctrl"
cache size :31457280
ok writing benchmark parameters to resctrl FS ok Write schema "L3:0=7fff" to resctrl FS ok cache difference more than 4 % # Percent diff=16 # Number of bits: 15 #Avg_llc_perf_miss: 306994 # Allocated cache lines: 368640 ok
CAT: test
1..11
=================Appreciate any feedback on this!
Cheers,
Andre.[1] http://linux-arm.org/git?p=linux-ap.git;a=shortlog;h=refs/heads/resctrl-
selftests-wipgit://linux-arm.org/linux-ap.gitbranch: resctrl-selftests-wip
James Morse
reviewerHi Fenghua,
Quoted 13 lines · Fenghua
Quoted 12 lines
On Friday, May 10, 2019 10:36 AM
Andre Przywara [mailto:[email protected]] wrote:Quoted 3 lines · On Sat, 9 Feb 2019 18:50:29 -0800 Fenghua Yu <[email protected]> wrote:
With more and more resctrl features are being added by Intel, AMD and ARM, a test tool is becoming more and more useful to validate that both hardware and software functionalities work as expected.
That's very much appreciated! We decided to use that tool here to detect regressions in James' upcoming resctrl rework series. While doing so we spotted some shortcomings:
Quoted 6 lines
Quoted 2 lines
- There is some unconditional x86 inline assembly which obviously breaks the build on ARM.
Will fix this as much as possible.
BTW, does ARM support perf
imc_countevents which are used in CAT tests?
I've never heard of these. git-grep says its a powerpc pmu...
(after a quick chat with Andre), is this a cache-miss counter?
If so, its a bit murky, (and beware, I don't know much about perf).
The arch code's armv8_pmu has a 'PERF_COUNT_HW_CACHE_MISSES' map entry, so yes... ... but if we're measuring this on a cache outside the CPU, we'd need an 'uncore' pmu driver, so it would depend on what the manufacturer implemented.
I should admit that I'm expecting this selftest to test resctrl, and not depend on a lot else. Couldn't it use the llc_occupancy value, 50% bitmap gives ~50% lower occupancy. Or (some combination of) the mbm byte counters, (which would require a seeky workload to cause repeat re-fills of the same line)...
Thanks,
James