genode/repos/gems/run/depot_autopilot.run
Martin Stein 3f60dcfae8 sel4 run/depot_autopilot: handle resource leak
After a certain number of tests, presumably some resource in core is exceeded
and loading the successive test fails. This quickfix looks out for the
characteristic Core error and then reboots to avoid that all successive tests
are marked as failed.
2019-01-14 12:34:41 +01:00

813 lines
23 KiB
Plaintext

#
# Integration of Depot Autopilot into a Run-tool-based test infrastructure
#
# For a detailed documentation of the user interface and the features of this
# Run script please see _repos/gems/src/app/depot_autopilot/README_.
#
##############################################################
## Local copies of run tool procedures with small adaptions ##
## ##
## FIXME: Adapt original and remove local copies ##
##############################################################
proc autopilot_wait_for_output { wait_for_re timeout_value running_spawn_id } {
global output
global run_genode_failed
if {$wait_for_re == "forever"} {
set timeout -1
interact {
\003 {
send_user "Expect: 'interact' received 'strg+c' and was cancelled\n";
exit
}
-i $running_spawn_id
}
} else {
set timeout $timeout_value
}
set platform_msg [run_boot_string]
if {$platform_msg eq ""} {
set platform_msg "undefined platform command startup string sequence"
}
expect {
-i $running_spawn_id $platform_msg { puts stderr "Error: platform rebooted unexpectedly"; exit -4 }
-i $running_spawn_id -re $wait_for_re { }
# sel4
-i $running_spawn_id -re {Error: ~Cnode - not implemented.*?\n} {
puts stderr "Error: Core presumably went out of resources";
set run_genode_failed 1
return
}
# pistachio kernel fault
-i $running_spawn_id -re {--- "KD# Exception caught" ---.*?\n} {
puts stderr "Error: Kernel fault";
set run_genode_failed 1
return
}
# sel4 unknown fault caught by core
-i $running_spawn_id -re {Error: unexpected exception during fault.*?stopped.*?\n} {
puts stderr "Error: Unknown fault";
set run_genode_failed 1
return
}
# can happen, for instance, on socat TCP-timeout
eof {
puts stderr "Error: Spawned process died unexpectedly";
set run_genode_failed 1
return
}
timeout {
puts stderr "Error: Test execution timed out";
set run_genode_failed 1
return
}
}
append output $expect_out(buffer)
}
proc autopilot_create_tar_from_depot_binaries { archive_path args } {
# filter out api and src archives from requested depot content
set content {}
foreach subdir [_collect_from_depot $args] {
if {[regexp [_depot_archive_versioned_path_pattern] $subdir dummy user type]} {
if {$type == "src"} continue;
if {$type == "api"} continue;
}
lappend content $subdir
}
check_for_missing_depot_archives
eval "exec tar cf $archive_path -T /dev/null -C [depot_dir] [lsort -unique $content]"
}
proc autopilot_run_genode_until {{wait_for_re forever} {timeout_value 0} {running_spawn_id -1}} {
#
# If a running_spawn_id is specified, wait for the expected output
#
if {$running_spawn_id != -1} {
autopilot_wait_for_output $wait_for_re $timeout_value $running_spawn_id
return;
}
set retry 3
while { $retry != 0 } {
if {[expr [run_power_on] == false]} {
puts "Power on step failed, retry."
sleep 3
incr retry -1;
continue
}
if {![run_load]} {
puts "Load step failed, retry."
# kill the spawned load process if there is one
if {[load_spawn_id] != -1} {
set pid [exp_pid -i [load_spawn_id]]
exec kill -9 $pid
}
incr retry -1;
continue;
}
if {![run_log $wait_for_re $timeout_value]} {
puts "Log step failed, retry."
incr retry -1;
continue;
}
return;
}
puts stderr "Boot process failed 3 times in series. I give up!";
exit -1;
}
########################
## Utility procedures ##
########################
proc depot_user {} { return [get_cmd_arg --depot-user genodelabs] }
#
# Get value of an environment variable
#
proc get_env_var { var_name default_value } {
if {[info exists ::env($var_name)]} {
return $::env($var_name)
}
return $default_value
}
#
# Check if archives are available without doing anything with them
#
proc check_archives_available { args } {
# filter out api and src archives from requested depot content
set content {}
foreach subdir [_collect_from_depot $args] {
if {[regexp [_depot_archive_versioned_path_pattern] $subdir dummy user type]} {
if {$type == "src"} continue;
if {$type == "api"} continue;
}
lappend content $subdir
}
check_for_missing_depot_archives
}
#
# Return routes for boot modules that shall overlay the test-depot content
#
proc single_test_module_routes { } {
global test_modules
set result ""
foreach module $test_modules {
append result {
<service name="ROM" label_last="} $module {"> <parent/> </service>}
}
return $result
}
#
# Return autopilot start-nodes for the test packages that shall be run
#
proc test_pkgs_start_nodes { } {
global test_pkgs
global skip_test_pkg
set result ""
foreach test_pkg $test_pkgs {
if {[info exists skip_test_pkg($test_pkg)] && $skip_test_pkg($test_pkg)} {
append result {
<start name="} $test_pkg {" skip="true"/>}
} else {
append result {
<start name="} $test_pkg {" pkg="} [depot_user] {/pkg/} $test_pkg {/} [_current_depot_archive_version pkg $test_pkg] {"/>}
}
}
return $result
}
#
# Prepare to call run_genode_until (again, with a changed setup)
#
proc prepare_to_run_genode { } {
global output
global qemu_args
global previous_results
global previous_time_ms
global previous_succeeded
global previous_failed
global previous_skipped
global skip_test_pkg
global test_pkgs
global test_srcs
global test_builds
global test_modules
global test_repeat
global last_test_pkg
global run_genode_failed
global serial_id
global timeout
global initial_qemu_args
set qemu_args $initial_qemu_args
#
# Create a depot archive that contains the test packages
#
create_boot_directory
set depot_tar_archives ""
set depot_tar_src_archives ""
set import_archives ""
foreach test_pkg $test_pkgs {
if { [info exists skip_test_pkg($test_pkg)] } {
append_if [expr !$skip_test_pkg($test_pkg)] depot_tar_archives " [depot_user]/pkg/$test_pkg "
} else {
append depot_tar_archives " [depot_user]/pkg/$test_pkg "
}
}
foreach test_src_pkg $test_srcs {
if { [info exists skip_test_pkg($test_src_pkg)] } {
append_if [expr !$skip_test_pkg($test_src_pkg)] depot_tar_src_archives " [depot_user]/src/$test_src_pkg "
} else {
append depot_tar_src_archives " [depot_user]/src/$test_src_pkg "
}
}
append import_archives {
} [depot_user] {/src/} [base_src] {
} [depot_user] {/src/report_rom
} [depot_user] {/src/fs_rom
} [depot_user] {/src/vfs
} [depot_user] {/src/loader
} [depot_user] {/src/init
} [depot_user] {/src/depot_query
}
if {!$skip_test_pkg(test-lx_block)} {
append import_archives [depot_user] {/raw/test-lx_block }
}
set all_archives [concat $depot_tar_archives $import_archives]
check_archives_available {*}$all_archives
autopilot_create_tar_from_depot_binaries [run_dir]/genode/depot.tar {*}$depot_tar_archives
set depot_tar_wo_src_size [file size [run_dir]/genode/depot.tar]
append_src_and_api_depot_packages_to_tar [run_dir]/genode/depot.tar {*}$depot_tar_src_archives
set depot_tar_size [file size [run_dir]/genode/depot.tar]
puts "Depot archive has a size of $depot_tar_size bytes ($depot_tar_wo_src_size bytes without sources)"
#
# Install the root-init config
#
append config {
<config prio_levels="2">
<parent-provides>
<service name="ROM"/>
<service name="IRQ"/>
<service name="IO_PORT"/>
<service name="IO_MEM"/>
<service name="PD"/>
<service name="RM"/>
<service name="CPU"/>
<service name="LOG"/>
<service name="TRACE"/>
</parent-provides>
<default-route>
<any-service> <parent/> <any-child/> </any-service>
</default-route>
<default caps="100"/>
<start name="timer" priority="0">
<resource name="RAM" quantum="1M"/>
<provides> <service name="Timer"/> </provides>
</start>
<start name="report_rom" priority="-1">
<binary name="report_rom"/>
<resource name="RAM" quantum="2M"/>
<provides> <service name="Report"/> <service name="ROM"/> </provides>
<config verbose="no">
<policy label="depot_autopilot -> blueprint" report="depot_query -> blueprint"/>
<policy label="depot_query -> query" report="depot_autopilot -> query"/>
<policy label="dynamic -> config" report="depot_autopilot -> init.config"/>
</config>
</start>
<start name="vfs" priority="-1">
<resource name="RAM" quantum="16M"/>
<provides> <service name="File_system"/> </provides>
<config>
<vfs>
<dir name="depot"> <tar name="depot.tar"/> </dir>
<dir name="gcov_data"> <ram/> </dir>
</vfs>
<policy label="depot_query -> depot" root="/depot" />
<policy label="fs_rom -> " root="/depot" />
<policy label="dynamic -> gcov -> gcov -> depot" root="/depot" />
<policy label_suffix=" -> gcov_data" root="/gcov_data" writeable="yes" />
</config>
</start>
<start name="fs_rom" priority="-1" caps="300">
<resource name="RAM" quantum="16M"/>
<provides> <service name="ROM"/> </provides>
</start>
<start name="depot_query" priority="-1">
<resource name="RAM" quantum="2M"/>
<config query="rom">
<vfs> <dir name="depot"> <fs label="depot"/> </dir> </vfs>
</config>
<route>
<service name="ROM" label="query"> <child name="report_rom"/> </service>
<any-service> <parent/> <any-child/> </any-service>
</route>
</start>
<start name="depot_autopilot" priority="-1">
<resource name="RAM" quantum="2M"/>
<provides> <service name="LOG"/> </provides>
<config repeat="} $test_repeat {" arch="} [depot_spec] {" children_label_prefix="dynamic -> ">
<static>
<parent-provides>
<service name="ROM"/>
<service name="CPU"/>
<service name="PD"/>
<service name="LOG"/>
<service name="RM"/>
<service name="Timer"/>
<service name="TRACE"/>
<service name="File_system"/>
</parent-provides>
</static>
<common_routes>
} [single_test_module_routes] {
<service name="ROM" label_last="ld.lib.so"> <parent/> </service>
<service name="ROM" label_last="init"> <parent/> </service>
<service name="CPU"> <parent/> </service>
<service name="PD"> <parent/> </service>
<service name="LOG"> <parent/> </service>
<service name="Timer"> <parent/> </service>
<service name="RM"> <parent/> </service>
<service name="TRACE"> <parent/> </service>
<service name="File_system" label="gcov -> depot"> <parent/> </service>
<service name="File_system" label_suffix="gcov_data"> <parent/> </service>
</common_routes>
<previous-results time_sec="} [expr ($previous_time_ms + 999) / 1000] {"
succeeded="} $previous_succeeded {"
skipped="} $previous_skipped {"
failed="} $previous_failed {">} [string map {\" .} $previous_results] {</previous-results>
} [test_pkgs_start_nodes] {
</config>
<route>
<service name="PD" unscoped_label="depot_autopilot"> <parent/> </service>
<service name="CPU" unscoped_label="depot_autopilot"> <parent/> </service>
<service name="LOG" unscoped_label="depot_autopilot"> <parent/> </service>
<service name="ROM" unscoped_label="depot_autopilot"> <parent/> </service>
<service name="ROM" unscoped_label="ld.lib.so"> <parent/> </service>
<service name="ROM" label="blueprint"> <child name="report_rom"/> </service>
<service name="Report" label="query"> <child name="report_rom"/> </service>
<service name="Report" label="init.config"> <child name="report_rom"/> </service>
<service name="Timer" label=""> <child name="timer"/> </service>
</route>
</start>
<start name="dynamic" caps="8000" priority="-1">
<resource name="RAM" quantum="420M"/>
<binary name="init"/>
<route>
} [single_test_module_routes] {
<service name="ROM" label_last="ld.lib.so"> <parent/> </service>
<service name="ROM" label_last="init"> <parent/> </service>
<service name="ROM" label="config"> <child name="report_rom"/> </service>
<service name="ROM"> <child name="fs_rom"/> </service>
<service name="LOG" unscoped_label="dynamic"> <parent/> </service>
<service name="LOG"> <child name="depot_autopilot"/> </service>
<service name="Timer"> <child name="timer"/> </service>
<any-service> <parent/> <any-child/> </any-service>
</route>
</start>
</config>}
install_config $config
#
# Create the rest of the boot modules
#
set build_components { app/depot_autopilot }
append build_components $test_builds
build $build_components
import_from_depot {*}$import_archives
#
# Build boot image from boot modules
#
set boot_modules { depot_autopilot }
append boot_modules $test_modules
build_boot_image $boot_modules
set last_test_pkg ""
set run_genode_failed 0
set serial_id -1
set timeout 40
append qemu_args "-nographic -serial mon:stdio "
}
#
# Initialize variables that accumulate test results for possible reboots
#
proc init_previous_results {} {
global previous_results_on_exit
global previous_results
global previous_time_ms
global previous_succeeded
global previous_failed
global previous_skipped
set previous_results_on_exit 1
set previous_results ""
set previous_time_ms 0
set previous_succeeded 0
set previous_failed 0
set previous_skipped 0
}
#
# Initialize variables that can be configured via the user interface
#
proc init_test_setting {} {
global test_pkgs
global test_srcs
global test_builds
global test_modules
global test_repeat
global default_test_pkgs
global default_test_srcs
global skip_test_pkg
#
# Initialize the lists of test package-archives, source-archives,
# build-targets, and boot-modules to be considered by the run script
#
set test_pkgs [get_env_var TEST_PKGS ""]
set test_srcs [get_env_var TEST_SRCS ""]
set test_builds [get_env_var TEST_BUILDS ""]
set test_modules [get_env_var TEST_MODULES ""]
set test_repeat [get_env_var TEST_REPEAT "false"]
#
# If the user didn't declare an individual list of test package-archives, use
# the default lists of test package-archives and source-archives
#
set nr_of_tests_to_run 0
if {$test_pkgs == ""} {
foreach test_pkg $default_test_pkgs {
append test_pkgs " $test_pkg "
if {!([info exists skip_test_pkg($test_pkg)] && $skip_test_pkg($test_pkg))} {
incr nr_of_tests_to_run
}
}
foreach test_src_pkg $default_test_srcs {
append test_srcs " $test_src_pkg "
}
} else {
foreach test_pkg $test_pkgs {
if {!([info exists skip_test_pkg($test_pkg)] && $skip_test_pkg($test_pkg))} {
incr nr_of_tests_to_run
}
}
}
}
rename exit run_tool_exit
proc exit {{status 0}} {
global previous_results_on_exit
global previous_results
global previous_time_ms
global previous_succeeded
global previous_failed
global previous_skipped
if {![info exists previous_results_on_exit] || \
![info exists previous_results] || \
![info exists previous_time_ms] || \
![info exists previous_succeeded] || \
![info exists previous_failed] || \
![info exists previous_skipped]} \
{
init_previous_results
}
if {$previous_results_on_exit != 0} {
set previous_time_sec [expr $previous_time_ms / 1000]
set previous_time_sec_frac [expr $previous_time_ms - $previous_time_sec * 1000]
set previous_results_list [split $previous_results "\n"]
puts ""
puts "Failed to let Depot Autopilot finish!"
puts "Result overview simulated by run script:"
puts ""
puts "\[init -> depot_autopilot] --- Finished after $previous_time_sec.$previous_time_sec_frac sec ---"
puts "\[init -> depot_autopilot] "
foreach previous_result $previous_results_list {
puts "\[init -> depot_autopilot] $previous_result"
}
puts "\[init -> depot_autopilot] "
puts "\[init -> depot_autopilot] succeeded: $previous_succeeded failed: $previous_failed skipped: $previous_skipped"
puts "\[init -> depot_autopilot] "
}
run_tool_exit $status
}
##################
## Main routine ##
##################
#
# Check platform support
#
if {[expr ![have_spec x86] && \
![have_spec arm_v6] && \
![have_spec arm_v7a]] || \
[expr [have_spec foc] && [have_spec panda]]} \
{
puts "\n Run script is not supported on this platform. \n";
exit 0
}
#
# Default list of test package-archives
#
# Was obtained by issuing:
#
# ! cd <GENODE_DIR>/repos
# ! find . -type d -wholename *recipes/pkg/test-* -printf '%f\n' | sort
#
set default_test_pkgs {
test-ada
test-ada_exception
test-ada_secondary_stack
test-block
test-block_cache
test-clipboard
test-ds_ownership
test-dynamic_config
test-dynamic_config_loader
test-dynamic_config_slave
test-expat
test-fault_detection
test-fs_log
test-fs_packet
test-fs_report
test-fs_rom_update
test-fs_rom_update_fs
test-fs_rom_update_ram
test-gnatio
test-init
test-init_loop
test-ldso
test-libc
test-libc_counter
test-libc_getenv
test-libc_pipe
test-libc_vfs
test-libc_vfs_audit
test-libc_vfs_block
test-libc_vfs_counter
test-libc_vfs_fs
test-libc_vfs_fs_chained
test-libc_vfs_ram
test-log
test-lx_block
test-magic_ring_buffer
test-mmio
test-new_delete
test-nic_loopback
test-part_block_gpt
test-part_block_mbr
test-pipe
test-pthread
test-python
test-ram_fs_chunk
test-read_only_rom
test-reconstructible
test-registry
test-report_rom
test-resource_request
test-resource_yield
test-rm_fault
test-rm_fault_no_nox
test-rm_nested
test-rom_block
test-rom_filter
test-rust
test-sanitizer
test-sequence
test-signal
test-slab
test-solo5
test-stdcxx
test-synced_interface
test-tcp_bulk_lwip
test-tcp_bulk_lxip
test-terminal_crosslink
test-timed_semaphore
test-timer
test-trace
test-trace_logger
test-utf8
test-vfs_stress_fs
test-vfs_stress_ram
test-weak_ptr
test-xml_generator
test-xml_node
gcov
}
#
# Default list of test source-archives
#
set default_test_srcs {
test-xml_generator
}
#
# Whether the platform supports non-executable dataspaces
#
proc non_executable_supported { } {
if {[have_spec hw] && [have_spec x86_64]} { return true }
if {[have_spec hw] && [have_spec arm]} { return true }
if {[have_spec nova] && [have_spec x86_64]} { return true }
if {[have_spec foc] && [have_spec x86_64]} { return true }
if {[have_spec foc] && [have_spec arm]} { return true }
if {[have_spec sel4] && [have_spec arm]} { return true }
return false
}
#
# Whether to skip a test - if undefined for a test, the test is not skipped
#
set skip_test_pkg(test-python) [expr ![have_spec x86]]
set skip_test_pkg(test-slab) [expr [get_cmd_switch --autopilot] && [have_include "power_on/qemu"]]
set skip_test_pkg(test-rm_nested) [expr [have_spec linux]]
set skip_test_pkg(test-fault_detection) [expr [have_spec pistachio] || [have_spec fiasco]]
set skip_test_pkg(test-fs_packet) [expr [get_cmd_switch --autopilot] && [have_include "power_on/qemu"]]
set skip_test_pkg(test-rm_fault) [expr [have_spec linux] || ![non_executable_supported]]
set skip_test_pkg(test-rm_fault_no_nox) [expr [have_spec linux] || !$skip_test_pkg(test-rm_fault)]
set skip_test_pkg(test-lx_block) [expr ![have_spec linux]]
set skip_test_pkg(test-tcp_bulk_lwip) [expr ![have_spec x86]]
set skip_test_pkg(test-tcp_bulk_lxip) [expr ![have_spec x86]]
set skip_test_pkg(test-solo5) [expr ![have_spec x86_64]]
set skip_test_pkg(test-libc) [expr [have_spec sel4] || [have_spec rpi] || [expr [have_spec pbxa9] && [have_spec foc]] || [expr [have_spec imx53] && [have_spec trustzone]]]
#
# FIXME
#
# When doing the libc_getenv test on autopilot+foc+x86 and one of the
# subsequent tests crashes the system so it gets rebooted by the run script,
# the system doesn't come up again. It gets stuck after core initialization.
#
set skip_test_pkg(test-libc_getenv) [expr [get_cmd_switch --autopilot] && [have_spec foc] && [have_spec x86]]
# remember initial qemu args in case we have to re-boot later
set initial_qemu_args ""
if {[info exists qemu_args]} {
set initial_qemu_args $qemu_args
}
# initialize global variables
init_test_setting
init_previous_results
# generic preparation for each system boot
prepare_to_run_genode
while {1} {
# wait for the next autopilot event
if {$serial_id == -1} {
autopilot_run_genode_until {depot_autopilot\] --- .*?\n} $timeout
set serial_id [output_spawn_id]
# if the system didn't even boot, exit (prints previous results)
if {$run_genode_failed} {
exit -1
}
} else {
set init_time_ms [clock clicks -millisec]
autopilot_run_genode_until {depot_autopilot\] --- .*?\n} $timeout $serial_id
set previous_time_ms [expr $previous_time_ms + [expr ([clock clicks -millisec] - $init_time_ms)] ]
set serial_id [output_spawn_id]
# remove last test from list and check if we have to reboot the system
set test_pkgs [lsearch -all -inline -not -exact $test_pkgs $last_test_pkg]
if {$run_genode_failed} {
# shut-down running system
exec kill -9 [exp_pid -i $serial_id]
run_power_off
# remember result of last test
if {$previous_results != ""} {
append previous_results \012
}
append previous_results { } [format {%-31s %-6s %7s} $last_test_pkg "failed " "$timeout.000"] { reboot}
incr previous_failed
# prepare system re-boot
prepare_to_run_genode
continue
}
}
# if the autopilot finished all tests, evaluate its return value
if {[regexp {depot_autopilot\] --- Finished} $output]} {
set output ""
run_genode_until {child "depot_autopilot" exited with exit value.*?\n} 10 $serial_id
set previous_results_on_exit 0
grep_output {^\[init\] }
compare_output_to {[init] child "depot_autopilot" exited with exit value 0}
exit 0
}
# if the autopilot started a new test, set a new timeout
if {[regexp {depot_autopilot\] --- Run} $output]} {
if {$last_test_pkg != ""} {
# remember result of last test in case the system must be restartet
set last_test_result ""
regexp {depot_autopilot\] ( [^\033]+)} $output ignored last_test_result
regsub -all {<} $last_test_result {\&lt;} last_test_result
set failed_off [string first " failed" $last_test_result]
set skipped_off [string first " skipped" $last_test_result]
set ok_off [string first " ok" $last_test_result]
if {$failed_off > 0 && ($skipped_off < 0 || $failed_off < $skipped_off) && ($ok_off < 0 || $failed_off < $ok_off)} {
incr previous_failed
} elseif {$skipped_off > 0 && ($ok_off < 0 || $skipped_off < $ok_off)} {
incr previous_skipped
} elseif {$ok_off > 0} {
incr previous_succeeded
} else {
puts "Error: malformed test result"
puts $last_test_result
exit -1
}
if {$previous_results != ""} {
append previous_results \012
}
append previous_results $last_test_result
}
# if the Autopilot is currently repeating, reset repeat-influenced variables
if {[llength $test_pkgs] == 0} {
init_test_setting
init_previous_results
}
# determine timeout for the next timeout
set min_timeout 10
regexp {depot_autopilot\] --- Run "(.*?)" \(max ([0-9]*?) } $output] ignore last_test_pkg min_timeout
set timeout [expr $min_timeout + 10]
set output ""
}
}