| 1 |
# An implementation of generating passlist for safe-probes |
| 2 |
# in systemtap based on the discussion in thread: |
| 3 |
# http://sourceware.org/ml/systemtap/2006-q3/msg00574.html |
| 4 |
# |
| 5 |
# Copyright (C) 2006 IBM Corp. |
| 6 |
# |
| 7 |
# This file is part of systemtap, and is free software. You can |
| 8 |
# redistribute it and/or modify it under the terms of the GNU General |
| 9 |
# Public License (GPL); either version 2, or (at your option) any |
| 10 |
# later version. |
| 11 |
# |
| 12 |
# Main ideas: |
| 13 |
# 1)Fetch a group of probe points from probes.pending, probe them |
| 14 |
# and run some workloads(e.g. runltp) parallely meanwhile. |
| 15 |
# |
| 16 |
# 2)If the probe test ends without crash, those actually triggered |
| 17 |
# probe points are moved into probes.passed and those |
| 18 |
# untriggered are into probes.untriggered; if the probe test |
| 19 |
# crashes the system, it will be resumed automatically after |
| 20 |
# system reboot. Those probe points which have been triggered are |
| 21 |
# also moved into probes.passed, but those untriggered ones are |
| 22 |
# moved into probes.failed. |
| 23 |
# |
| 24 |
# 3)Repeat the above until probes.pending becomes empty, then: |
| 25 |
# Normally, probes.pending is reinitialized from probes.failed |
| 26 |
# and probes.untriggered, then start the next iteration; |
| 27 |
# But if max running level (now 3) is reached, or probes.pending, |
| 28 |
# probes.failed and probes.untriggered are all empty, stop the |
| 29 |
# whole test. |
| 30 |
# |
| 31 |
# http://sourceware.org/ml/systemtap/2006-q4/msg00143.html |
| 32 |
# http://sourceware.org/ml/systemtap/2006-q4/msg00435.html |
| 33 |
# |
| 34 |
set TESTNAME "passlist" |
| 35 |
|
| 36 |
# Because this test will take several hours (depends on workloads |
| 37 |
# and max_running_level) and probably cause more than once system |
| 38 |
# crashes, it is disabled by default. It's strongly suggested |
| 39 |
# to run it manually as a single test case if necessary. |
| 40 |
# Comment the following two lines to enable it. |
| 41 |
untested "$TESTNAME is disabled" |
| 42 |
return |
| 43 |
|
| 44 |
################################################################# |
| 45 |
# supporting constants and procedures # |
| 46 |
################################################################# |
| 47 |
set MAX_RUNNING_LEVEL 2 |
| 48 |
|
| 49 |
set STP_GENWHITELIST_RUNNING "/stp_genpasslist_running" |
| 50 |
set STAP_RESULT "probe.out" |
| 51 |
set PROBES_ALL "probes.all" |
| 52 |
set PROBES_PENDING "probes.pending" |
| 53 |
set PROBES_CURRENT "probes.current" |
| 54 |
set PROBES_PASSED "probes.passed" |
| 55 |
set PROBES_FAILED "probes.failed" |
| 56 |
set PROBES_UNTRIGGERED "probes.untriggered" |
| 57 |
|
| 58 |
set systemtap_script { |
| 59 |
global stat |
| 60 |
probe %s { |
| 61 |
stat[pp()] <<< 1 |
| 62 |
} |
| 63 |
probe timer.ms(30000) { |
| 64 |
foreach (pp in stat) |
| 65 |
printf("%%d %%s\n", @count(stat[pp]), pp) |
| 66 |
} |
| 67 |
probe end { |
| 68 |
foreach (pp in stat) |
| 69 |
printf("%%d %%s\n", @count(stat[pp]), pp) |
| 70 |
} |
| 71 |
} |
| 72 |
|
| 73 |
set benchs { |
| 74 |
{/usr/local/ltp/runltp -t 300s -f syscalls} |
| 75 |
{/usr/local/ltp/runltp -t 300s -f nfs} |
| 76 |
{/usr/local/ltp/runltp -t 300s -f ipc} |
| 77 |
{/usr/local/ltp/runltp -t 300s -f dio} |
| 78 |
{/usr/local/ltp/runltp -t 300s -f fs} |
| 79 |
{/usr/local/ltp/runltp -t 300s -f mm} |
| 80 |
{/usr/local/ltp/runltp -t 300s -f tcp_cmds} |
| 81 |
{/usr/local/ltp/runltp -t 300s -f ltp-aio-stress.part1} |
| 82 |
} |
| 83 |
|
| 84 |
set init_probes_all_script { |
| 85 |
probe scheduler.*, |
| 86 |
ioscheduler.*, |
| 87 |
ioscheduler.*.return, |
| 88 |
syscall.*, |
| 89 |
syscall.*.return, |
| 90 |
ioblock.*, |
| 91 |
netdev.*, |
| 92 |
vm.*, |
| 93 |
signal.*, |
| 94 |
signal.*.return, |
| 95 |
udp.*, |
| 96 |
udp.*.return, |
| 97 |
tcp.*, |
| 98 |
tcp.*.return, |
| 99 |
kprocess.*, |
| 100 |
process.*, |
| 101 |
nfs.fop.*, |
| 102 |
nfs.aop.*, |
| 103 |
nfsd.proc.*, |
| 104 |
nfsd.*, |
| 105 |
nfs.proc.*, |
| 106 |
scsi.*, |
| 107 |
sunrpc.*, |
| 108 |
generic.fop.*, |
| 109 |
generic.fop.*.return, |
| 110 |
vfs.*, |
| 111 |
vfs.*.return {} |
| 112 |
} |
| 113 |
|
| 114 |
proc do_current_test {} { |
| 115 |
global PROBES_CURRENT |
| 116 |
global STAP_RESULT |
| 117 |
global systemtap_script |
| 118 |
if [catch {open $PROBES_CURRENT r} Infile] { |
| 119 |
puts "Failed to open $PROBES_CURRENT" |
| 120 |
return |
| 121 |
} |
| 122 |
set probepoint [gets $Infile] |
| 123 |
while {![eof $Infile]} { |
| 124 |
set extra [gets $Infile] |
| 125 |
if {[string length $extra] > 0} { |
| 126 |
set probepoint [concat "$probepoint, " $extra] |
| 127 |
set extra [gets $Infile] |
| 128 |
} |
| 129 |
} |
| 130 |
catch {close $Infile} |
| 131 |
|
| 132 |
set testname "L[get_running_level]_Grp_[string range $probepoint 0 32]..." |
| 133 |
if [probe_ok $probepoint] { |
| 134 |
set script [format $systemtap_script $probepoint] |
| 135 |
passlist_run $testname runbenchs -e $script -o $STAP_RESULT |
| 136 |
} else { |
| 137 |
puts "ERROR no match in probe_ok() $testname no match" |
| 138 |
set script [format $systemtap_script $probepoint] |
| 139 |
exec echo $script > sc.stp |
| 140 |
} |
| 141 |
} |
| 142 |
|
| 143 |
proc get_linesize {filename} { |
| 144 |
if [file readable $filename] { |
| 145 |
scan [exec wc -l $filename] "%d" lines |
| 146 |
if {[info exists lines] && $lines > 0} { |
| 147 |
return $lines |
| 148 |
} |
| 149 |
} |
| 150 |
return 0 |
| 151 |
} |
| 152 |
|
| 153 |
proc get_running_level {} { |
| 154 |
global STP_GENWHITELIST_RUNNING |
| 155 |
if [file readable $STP_GENWHITELIST_RUNNING] { |
| 156 |
scan [exec cat $STP_GENWHITELIST_RUNNING] "%d" current_running_level |
| 157 |
if {[info exists current_running_level] && $current_running_level > 0} { |
| 158 |
return $current_running_level |
| 159 |
} |
| 160 |
} |
| 161 |
return 0 |
| 162 |
} |
| 163 |
|
| 164 |
set NO_CRASH 0 |
| 165 |
set ALREADY_CRASHED 1 |
| 166 |
proc garbage_collect {{already_crashed $NO_CRASH}} { |
| 167 |
global STAP_RESULT |
| 168 |
global PROBES_CURRENT |
| 169 |
global PROBES_PASSED |
| 170 |
global PROBES_FAILED |
| 171 |
global PROBES_UNTRIGGERED |
| 172 |
|
| 173 |
if {[get_linesize $PROBES_CURRENT] == 0} { |
| 174 |
exec rm -f $PROBES_CURRENT $STAP_RESULT |
| 175 |
return |
| 176 |
} |
| 177 |
if {[get_linesize $STAP_RESULT] == 0} { |
| 178 |
if {$already_crashed} { |
| 179 |
exec cat $PROBES_CURRENT >> $PROBES_FAILED |
| 180 |
} else { |
| 181 |
exec cat $PROBES_CURRENT >> $PROBES_UNTRIGGERED |
| 182 |
} |
| 183 |
exec rm -f $PROBES_CURRENT $STAP_RESULT |
| 184 |
return |
| 185 |
} |
| 186 |
|
| 187 |
# both probes.current and probe.out are non-empty |
| 188 |
if [catch {open $PROBES_CURRENT r} Infile] { |
| 189 |
puts "Failed to open $PROBES_CURRENT" |
| 190 |
exec rm -f $PROBES_CURRENT $STAP_RESULT |
| 191 |
return |
| 192 |
} |
| 193 |
set probepoint [gets $Infile] |
| 194 |
while {![eof $Infile]} { |
| 195 |
if {[is_probed $probepoint $STAP_RESULT]} { |
| 196 |
exec echo $probepoint >> $PROBES_PASSED |
| 197 |
} else { |
| 198 |
if {$already_crashed} { |
| 199 |
exec echo $probepoint >> $PROBES_FAILED |
| 200 |
} else { |
| 201 |
exec echo $probepoint >> $PROBES_UNTRIGGERED |
| 202 |
} |
| 203 |
} |
| 204 |
set probepoint [gets $Infile] |
| 205 |
} |
| 206 |
catch {close $Infile} |
| 207 |
exec rm -f $PROBES_CURRENT $STAP_RESULT |
| 208 |
return |
| 209 |
} |
| 210 |
|
| 211 |
proc incr_running_level {} { |
| 212 |
global STP_GENWHITELIST_RUNNING |
| 213 |
set newlevel [expr [get_running_level]+1] |
| 214 |
if { $newlevel > 0 } { |
| 215 |
exec echo $newlevel > $STP_GENWHITELIST_RUNNING |
| 216 |
} else { |
| 217 |
exec echo 0 > $STP_GENWHITELIST_RUNNING |
| 218 |
} |
| 219 |
return |
| 220 |
} |
| 221 |
|
| 222 |
proc init_probes_all {} { |
| 223 |
global PROBES_ALL |
| 224 |
global init_probes_all_script |
| 225 |
catch {exec stap -p2 -e $init_probes_all_script > /tmp/passlist_tmpfile} |
| 226 |
catch {exec grep "^kernel.function" /tmp/passlist_tmpfile > $PROBES_ALL } |
| 227 |
catch {exec rm -f /tmp/passlist_tmpfile} |
| 228 |
if {[get_linesize $PROBES_ALL] == 0} { |
| 229 |
return 1 |
| 230 |
} |
| 231 |
return 0 |
| 232 |
} |
| 233 |
|
| 234 |
proc init_running_level {} { |
| 235 |
global STP_GENWHITELIST_RUNNING |
| 236 |
exec echo 1 > $STP_GENWHITELIST_RUNNING |
| 237 |
} |
| 238 |
|
| 239 |
proc is_probed {probepoint stap_result} { |
| 240 |
if {[string length $probepoint] == 0 || [get_linesize $stap_result] == 0 } { |
| 241 |
return 0 |
| 242 |
} |
| 243 |
if [catch {open $stap_result r} Resfile] { |
| 244 |
puts "Failed to open $stap_result in is_probed() proc" |
| 245 |
return 0 |
| 246 |
} |
| 247 |
set probed 0 |
| 248 |
set resline [gets $Resfile] |
| 249 |
while {![eof $Resfile]} { |
| 250 |
if {[string match "* $probepoint" $resline]} { |
| 251 |
set probed 1 |
| 252 |
break |
| 253 |
} |
| 254 |
set resline [gets $Resfile] |
| 255 |
} |
| 256 |
catch {close $Resfile} |
| 257 |
return $probed |
| 258 |
} |
| 259 |
|
| 260 |
proc probe_ok {probepoint} { |
| 261 |
set cmd {exec stap -u -p2 -e } |
| 262 |
lappend cmd "probe $probepoint {}" |
| 263 |
exec echo $cmd > cmdfile |
| 264 |
return ![catch $cmd] |
| 265 |
} |
| 266 |
|
| 267 |
proc proper_current_size {level inputfile} { |
| 268 |
set totalsize [get_linesize $inputfile] |
| 269 |
switch $level { |
| 270 |
{1} {set currentsize [expr $totalsize/9]} |
| 271 |
{2} {set currentsize [expr $totalsize/49]} |
| 272 |
{3} {set currentsize 50} |
| 273 |
{4} {set currentsize 10} |
| 274 |
default { |
| 275 |
puts "Reached unexpected iteration level: $level" |
| 276 |
set currentsize $totalsize |
| 277 |
} |
| 278 |
} |
| 279 |
if {$currentsize <= 0} { |
| 280 |
set currentsize 5 |
| 281 |
} |
| 282 |
return $currentsize |
| 283 |
} |
| 284 |
|
| 285 |
set startup_line_in_RCLOCAL "cd $env(PWD); runtest passlist.exp&" |
| 286 |
set RCLOCAL "/etc/rc.d/rc.local" |
| 287 |
|
| 288 |
proc register_service {} { |
| 289 |
global startup_line_in_RCLOCAL |
| 290 |
global RCLOCAL |
| 291 |
exec sed -i -n -e "/runtest passlist.exp/!p" $RCLOCAL |
| 292 |
exec echo $startup_line_in_RCLOCAL >> $RCLOCAL |
| 293 |
} |
| 294 |
|
| 295 |
proc unregister_service {} { |
| 296 |
global RCLOCAL |
| 297 |
exec sed -i -n -e "/runtest passlist.exp/!p" $RCLOCAL |
| 298 |
} |
| 299 |
|
| 300 |
proc passlist_run { TEST_NAME {LOAD_GEN_FUNCTION ""} args } { |
| 301 |
set cmd [concat {stap -DMAXSKIPPED=200000 -v } $args] |
| 302 |
catch {eval spawn $cmd} |
| 303 |
set stap_id $spawn_id |
| 304 |
set failed 1 |
| 305 |
expect { |
| 306 |
-timeout 1800 |
| 307 |
-i $stap_id -re {^Pass\ ([1234]):\ [^\r]*\r\n} { |
| 308 |
set error_msg "pass$expect_out(1,string)"; |
| 309 |
exp_continue |
| 310 |
} |
| 311 |
-re {Pass\ 5:\ starting\ run.\r\n} { |
| 312 |
set error_msg "stap runtime" |
| 313 |
runbenchs |
| 314 |
kill -INT -[exp_pid -i $stap_id] |
| 315 |
exp_continue |
| 316 |
} |
| 317 |
-re {Pass\ 5:\ run\ completed} { |
| 318 |
set failed 0 |
| 319 |
} |
| 320 |
-re {parse\ error|semantic\ error} { set detail "$expect_out(0,string)" } |
| 321 |
timeout { set detail "stap timeout"; kill -INT -[exp_pid -i $stap_id] } |
| 322 |
eof { set failed 0 } |
| 323 |
} |
| 324 |
catch {close -i $stap_id} |
| 325 |
wait -i $stap_id |
| 326 |
if {$failed} { puts "passlist_run failure \($detail\)" } |
| 327 |
} |
| 328 |
|
| 329 |
proc runbenchs {} { |
| 330 |
global benchs |
| 331 |
set runningcount 0 |
| 332 |
|
| 333 |
foreach bench $benchs { |
| 334 |
set benchexec [lrange $bench 0 0] |
| 335 |
set benchargs [join [lrange $bench 1 [llength $bench]]] |
| 336 |
if {[file executable $benchexec]} { |
| 337 |
catch {eval spawn $benchexec $benchargs} |
| 338 |
set benchname($spawn_id) $benchexec |
| 339 |
} else { |
| 340 |
catch {eval spawn sleep 30} |
| 341 |
set benchname($spawn_id) "sleep" |
| 342 |
} |
| 343 |
lappend idlist $spawn_id |
| 344 |
incr runningcount |
| 345 |
} |
| 346 |
|
| 347 |
while {$runningcount > 0} { |
| 348 |
expect { |
| 349 |
-timeout 900 |
| 350 |
-i $idlist -re {LTP\ Version:\ LTP-([0-9])+\r\n$} { |
| 351 |
set from $expect_out(spawn_id) |
| 352 |
lappend benchres($from) $expect_out(buffer) |
| 353 |
} eof { |
| 354 |
set donepos [lsearch $idlist $expect_out(spawn_id)] |
| 355 |
set idlist [lreplace $idlist $donepos $donepos] |
| 356 |
incr runningcount -1 |
| 357 |
catch { close -i $expect_out(spawn_id) } |
| 358 |
wait -i $expect_out(spawn_id) |
| 359 |
} timeout { |
| 360 |
break |
| 361 |
} |
| 362 |
} |
| 363 |
} |
| 364 |
} |
| 365 |
|
| 366 |
################################################################### |
| 367 |
# Main routine of the whole test # |
| 368 |
################################################################### |
| 369 |
if {[info procs installtest_p] != "" && ![installtest_p]} { |
| 370 |
untested $TESTNAME; |
| 371 |
return |
| 372 |
} |
| 373 |
|
| 374 |
if {[get_running_level] == 0} { |
| 375 |
# Not started yet, start the whole test from the scratch |
| 376 |
# Append the startup code to /etc/rc.d/rc.local if not yet |
| 377 |
register_service |
| 378 |
# Check whether probes.all is empty or not given |
| 379 |
if {[get_linesize $PROBES_ALL] == 0} { |
| 380 |
if {[init_probes_all] != 0} { |
| 381 |
fail "$TESTNAME unable to initialize the probe point list" |
| 382 |
return |
| 383 |
} |
| 384 |
} |
| 385 |
# Set current_running_level as 1 to indicate a new test started |
| 386 |
init_running_level |
| 387 |
# Initialize intermediate files based on probe.all |
| 388 |
exec rm -f $PROBES_PENDING $PROBES_CURRENT |
| 389 |
exec rm -f $PROBES_PASSED $PROBES_UNTRIGGERED $PROBES_FAILED |
| 390 |
file copy $PROBES_ALL $PROBES_PENDING |
| 391 |
exec touch $PROBES_PASSED $PROBES_UNTRIGGERED $PROBES_FAILED |
| 392 |
puts "Start a fresh stp_genpasslist test." |
| 393 |
} else { |
| 394 |
# Maybe started already, so do some cleaning if necessary |
| 395 |
garbage_collect $ALREADY_CRASHED |
| 396 |
puts "Recovered from last maybe crashed probe test." |
| 397 |
} |
| 398 |
|
| 399 |
set current_size_const [proper_current_size [get_running_level] $PROBES_ALL] |
| 400 |
puts "current_size_const is initialized as $current_size_const" |
| 401 |
|
| 402 |
while {1} { |
| 403 |
puts "Current size of probes.pending is [get_linesize $PROBES_PENDING]" |
| 404 |
if {[get_linesize $PROBES_PENDING] == 0} { |
| 405 |
# Check whether we need the next iteration or not |
| 406 |
global MAX_RUNNING_LEVEL |
| 407 |
# incr running_level for the start of a new iteration |
| 408 |
incr_running_level |
| 409 |
puts "Running level increased to [get_running_level]" |
| 410 |
if {[get_running_level] > $MAX_RUNNING_LEVEL} { |
| 411 |
puts "Exceed max running level limit." |
| 412 |
break |
| 413 |
} else { |
| 414 |
puts "Current running level is [get_running_level]" |
| 415 |
exec rm -f $PROBES_PENDING |
| 416 |
if {[get_linesize $PROBES_FAILED] > 0} { |
| 417 |
# Append probes.failed to probes.pending |
| 418 |
exec cat $PROBES_FAILED >> $PROBES_PENDING |
| 419 |
file delete $PROBES_FAILED |
| 420 |
exec touch $PROBES_FAILED |
| 421 |
puts "Append $PROBES_FAILED to $PROBES_PENDING" |
| 422 |
} |
| 423 |
if {[get_linesize $PROBES_UNTRIGGERED] > 0} { |
| 424 |
# Append probes.untriggered to probes.pending |
| 425 |
exec cat $PROBES_UNTRIGGERED >> $PROBES_PENDING |
| 426 |
file delete $PROBES_UNTRIGGERED |
| 427 |
exec touch $PROBES_UNTRIGGERED |
| 428 |
puts "Append $PROBES_UNTRIGGERED to $PROBES_PENDING" |
| 429 |
} |
| 430 |
if {[get_linesize $PROBES_PENDING] == 0} { |
| 431 |
# No more pending probe points |
| 432 |
puts "No more iterations needed. Stopped." |
| 433 |
break |
| 434 |
} |
| 435 |
# set new value of current_size_const for new iteration level |
| 436 |
set current_size_const [proper_current_size [get_running_level] $PROBES_ALL] |
| 437 |
puts "current_size_const is set as $current_size_const now" |
| 438 |
continue |
| 439 |
} |
| 440 |
} |
| 441 |
# Now, non-empty probes.pending should be ready |
| 442 |
# Generate probes.current |
| 443 |
exec rm -f $PROBES_CURRENT |
| 444 |
exec head -n $current_size_const $PROBES_PENDING > $PROBES_CURRENT |
| 445 |
exec tail -n+[expr $current_size_const+1] $PROBES_PENDING > /tmp/passlist_tmpfile |
| 446 |
exec mv /tmp/passlist_tmpfile $PROBES_PENDING |
| 447 |
|
| 448 |
puts "Start a probe test..." |
| 449 |
# Do actual probe test |
| 450 |
do_current_test |
| 451 |
puts "Completed one probe test." |
| 452 |
|
| 453 |
# No crash fortunately, so do some cleaning to prepare for next test |
| 454 |
garbage_collect $NO_CRASH |
| 455 |
} |
| 456 |
|
| 457 |
# Congratulations for arriving here |
| 458 |
# Remove all temporary files and unregister myself |
| 459 |
puts "Remove all temporary files, unregister the service and return." |
| 460 |
exec rm -f $PROBES_PENDING $PROBES_CURRENT $STP_GENWHITELIST_RUNNING |
| 461 |
exec rm -f $STAP_RESULT /tmp/passlist_tmpfile |
| 462 |
unregister_service |
| 463 |
|
| 464 |
pass "$TESTNAME completed" |
| 465 |
return |