b=23182 lustre-iokit sgpdd-survey defect

[fs/lustre-release.git] / lustre-iokit / sgpdd-survey / sgpdd-survey

diff --git a/lustre-iokit/sgpdd-survey/sgpdd-survey b/lustre-iokit/sgpdd-survey/sgpdd-survey
index a46e92c..6d99457 100755 (executable)
--- a/lustre-iokit/sgpdd-survey/sgpdd-survey
+++ b/lustre-iokit/sgpdd-survey/sgpdd-survey
@@ -3,54 +3,103 @@
 ######################################################################
 # customize per survey
 
-# the SCSI devices to measure
-scsidevs="/dev/sde /dev/sdh"
+# CHOOSE EITHER scsidevs or rawdevs
+# the SCSI devices to measure - WARNING: will be erased.
+# The raw devices to use
+# rawdevs=${rawdevs:-"/dev/raw/raw1"}
+# scsidevs=`ls /dev/sd[a-z] /dev/sd[a-z][a-z]` # all devices, if you use udev
 
 # result file prefix.  date/time+hostname makes unique
 # NB ensure the path exists if it includes subdirs
-rslt=/tmp/sgpdd_survey_`date +%F@%R`_`uname -n`
+rslt_loc=${rslt_loc:-"/tmp"}
+rslt=${rslt:-"$rslt_loc/sgpdd_survey_`date +%F@%R`_`uname -n`"}
 
 # what to do (read or write)
-actions="write read"
+actions=${actions:-"write read"}
 
 # total size per device (MBytes)
 # NB bigger than device cache is good
-size=8192
+size=${size:-8192}
 
 # record size (KBytes)
-rszlo=1024
-rszhi=1024
+rszlo=${rszlo:-1024}
+rszhi=${rszhi:-1024}
 
 # Concurrent regions per device
-crglo=1
-crghi=256
+crglo=${crglo:-1}
+crghi=${crghi:-256}
 
 # threads to share between concurrent regions per device
+# multiple threads per region simulates a deeper request queue
 # NB survey skips over #thr < #regions and #thr/#regions > SG_MAX_QUEUE
-thrlo=1
-thrhi=4096
+thrlo=${thrlo:-1}
+thrhi=${thrhi:-4096}
 
 #####################################################################
 # leave the rest of this alone unless you know what you're doing...
 
-# sgp_dd's idea of disk sector size (Bytes)
-bs=512
 # and max # threads one instance will spawn
 SG_MAX_QUEUE=16
 
+# is the sg module loaded?
+sg_is_loaded=$(grep -q "^sg " /proc/modules && echo true || echo false)
+
+# did we load it?
+sg_was_loaded=false
+
 # map given device names into SG device names
 i=0
 devs=()
-for d in $scsidevs; do
-    devs[$i]=`sg_map | awk "{if ($ 2 == \"$d\") print $ 1}"`
-    if [ -z "$devs[$i]" ]; then
-       echo "Can't find SG device for $d"
-       exit 1
+if [ "$scsidevs" ]; then
+        # we will test for a LUN, the test for a partition
+        # if the partition number is > 9 this will fail
+
+    # make sure sg kernel module is loaded
+    if ! $sg_is_loaded; then
+       echo "loading the sg kernel module"
+       modprobe sg && sg_was_loaded=true
+       sg_is_loaded=true
     fi
-    i=$((i+1))
-done
+
+    for d in $scsidevs; do
+       devs[$i]=`sg_map | awk "{if (\\\$2 == \"$d\") print \\\$1}"`
+        if [ -z "${devs[i]}" ]; then
+            echo "Can't find SG device for $d, testing for partition"
+            pt=`echo $d | sed 's/[0-9]*$//'`
+            # Try again
+            devs[$i]=`sg_map | awk "{if (\\\$2 == \"$pt\") print \\\$1}"`
+            if [ -z "${devs[i]}" ]; then
+                echo -e "Can't find SG device $pt.\nDo you have the sg module configured for your kernel?"
+                exit 1
+           fi
+       fi
+       i=$((i+1))
+    done
+elif [ "$rawdevs" ]; then
+    for r in $rawdevs; do
+       RES=`raw -q $r`
+       if [ $? -eq 0 ];then
+           devs[$i]=$r
+           i=$((i+1))
+       else
+           echo "Raw device $r not set up"
+           exit 1
+       fi
+    done
+else
+    echo "Must specify scsidevs or rawdevs"
+    exit 1
+fi
+
 ndevs=${#devs[@]}
 
+# determine block size. This should also work for raw devices
+# If it fails, set to 512
+bs=$((`sg_readcap -b ${devs[0]} | awk '{print $2}'`))
+if [ $bs == 0  ];then
+       echo "sg_readcap failed, setting block size to 512"
+       bs=512
+fi
 rsltf=${rslt}.summary
 workf=${rslt}.detail
 echo -n > $rsltf
@@ -66,6 +115,8 @@ print_summary () {
     echo $minusn "$*"
 }
 
+print_summary "$(date) sgpdd-survey on $rawdevs$scsidevs from $(hostname)"
+
 for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do
     for ((crg=$crglo;crg<=$crghi;crg*=2)); do 
        for ((thr=$thrlo;thr<=$thrhi;thr*=2)); do
@@ -97,6 +148,13 @@ for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do
                t0=`date +%s.%N`
                for ((i=0;i<ndevs;i++)); do
                    dev=${devs[i]}
+                   devsize=$((bs*`sg_readcap -b ${dev} | awk '{print $1}'`/1024))
+                   if [ $devsize -lt $actual_size ]; then
+                       _dev=$(sg_map | grep $dev | awk '{ print $2; }')
+                       echo -e "device $_dev not big enough: $devsize <" \
+                               "$actual_size.\nConsider reducing \$size"
+                       exit 1
+                   fi
                    if [ $action = read ]; then
                        inf="if=$dev"
                        outf="of=/dev/null"
@@ -120,7 +178,9 @@ for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do
                for ((i=0;i<ndevs;i++)); do
                    for ((j=0;j<crg;j++)); do
                        rtmp=${tmpf}_${i}_${j}
-                       if grep 'time to transfer data' $rtmp > /dev/null 2>&1; then
+                       if grep 'error' $rtmp > /dev/null 2>&1; then
+                               echo "Error found in $rtmp"
+                       elif grep 'time to transfer data' $rtmp > /dev/null 2>&1; then
                            ok=$((ok + 1))
                        fi
                        cat ${rtmp} >> $tmpf
@@ -145,3 +205,8 @@ for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do
        done
     done
 done
+
+if $sg_was_loaded; then
+    echo "unloading sg module"
+    rmmod sg
+fi