Add support for idletime profiling

Idletime profiling allows a benchmark to run while filling the
idle cycles on the server, hence giving you some metric of how
much pressure the system is under. This is useful to be able
to profile and compare changes in a setup or driver, for instance.

Signed-off-by: Jens Axboe <[email protected]>

HOWTO

@@ -1693,3 +1693,18 @@ write      Write 'length' bytes beginning from 'offset'
 sync       fsync() the file
 datasync   fdatasync() the file
 trim       trim the given file from the given 'offset' for 'length' bytes
+
+
+9.0 CPU idleness profiling
+
+In some cases, we want to understand CPU overhead in a test. For example,
+we test patches for the specific goodness of whether they reduce CPU usage.
+fio implements a balloon approach to create a thread per CPU that runs at
+idle priority, meaning that it only runs when nobody else needs the cpu.
+By measuring the amount of work completed by the thread, idleness of each
+CPU can be derived accordingly.
+
+An unit work is defined as touching a full page of unsigned characters. Mean
+and standard deviation of time to complete an unit work is reported in "unit
+work" section. Options can be chosen to report detailed percpu idleness or
+overall system idleness by aggregating percpu stats.

Makefile

@@ -28,7 +28,8 @@ SOURCE := gettime.c fio.c ioengines.c init.c stat.c log.c time.c filesetup.c \
 		engines/mmap.c engines/sync.c engines/null.c engines/net.c \
 		memalign.c server.c client.c iolog.c backend.c libfio.c flow.c \
 		json.c lib/zipf.c lib/axmap.c lib/lfsr.c gettime-thread.c \
-		helpers.c lib/flist_sort.c lib/hweight.c lib/getrusage.c
+		helpers.c lib/flist_sort.c lib/hweight.c lib/getrusage.c \
+		idletime.c
 
 ifdef CONFIG_64BIT_LLP64
   CFLAGS += -DBITS_PER_LONG=32

README

@@ -145,6 +145,9 @@ $ fio
 	--max-jobs		Maximum number of threads/processes to support
 	--server=args		Start backend server. See Client/Server section.
 	--client=host		Connect to specified backend.
+	--idle-prof=option	Report cpu idleness on a system or percpu basis
+				(option=system,percpu) or run unit work
+				calibration only (option=calibrate).
 
 
 Any parameters following the options will be assumed to be job files,

backend.c

@@ -51,6 +51,7 @@
 #include "memalign.h"
 #include "server.h"
 #include "lib/getrusage.h"
+#include "idletime.h"
 
 static pthread_t disk_util_thread;
 static struct fio_mutex *disk_thread_mutex;
@@ -1499,6 +1500,8 @@ static void run_threads(void)
 
 	if (fio_gtod_offload && fio_start_gtod_thread())
 		return;
+
+	fio_idle_prof_init();
 
 	set_sig_handlers();
 
@@ -1556,6 +1559,9 @@ static void run_threads(void)
 		}
 	}
 
+	/* start idle threads before io threads start to run */
+	fio_idle_prof_start();
+
 	set_genesis_time();
 
 	while (todo) {
@@ -1718,6 +1724,8 @@ static void run_threads(void)
 			usleep(10000);
 	}
 
+	fio_idle_prof_stop();
+
 	update_io_ticks();
 	fio_unpin_memory();
 }

fio.1

@@ -77,6 +77,9 @@ Background a fio server, writing the pid to the given pid file.
 .TP
 .BI \-\-client \fR=\fPhost
 Instead of running the jobs locally, send and run them on the given host.
+.TP
+.BI \-\-idle\-prof \fR=\fPoption
+Report cpu idleness on a system or percpu basis (\fIoption\fP=system,percpu) or run unit work calibration only (\fIoption\fP=calibrate).
 .SH "JOB FILE FORMAT"
 Job files are in `ini' format. They consist of one or more
 job definitions, which begin with a job name in square brackets and