1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
|
# Copyright 2022 syzkaller project authors. All rights reserved.
# Use of this source code is governed by Apache 2 LICENSE that can be found in the LICENSE file.
config:
- CMDLINE: [append, "smp.csd_lock_timeout=100000 watchdog_thresh=55 workqueue.watchdog_thresh=140 sysctl.net.core.netdev_unregister_timeout_secs=140 dummy_hcd.num=32 max_loop=32 nbds_max=32"]
# Deterministic hang/stall detection.
# Without careful tuning the same hang/stall can be detected in multiple, radically-different ways.
# For example, an infinite loop can be detected as CPU stall, RCU stall, hung task or workqueue stall.
# This may lead to lots of duplicate bugs. In order to prevent this we strictly order hang/stall detection
# timeouts based on their causality (CPU stall can cause hung task, but not vise versa). See #516 for details.
#
# 1. RCU stalls has the highest priority.
# RCU_CPU_STALL_TIMEOUT=100 results in stalls detected after 100-101 secs.
# This also includes CSD lock timeout (CONFIG_CSD_LOCK_WAIT_DEBUG + smp.csd_lock_timeout=100000(ms)).
#
# 2. Then softlockup detector.
# kernel.watchdog_thresh = 55 detects stalls after 110-132 secs.
#
# 3. Then hung tasks, workqueue stalls and netdev unregister hangs.
# Unfortunately we can't separate them because that would require setting "no output" timeout to 10+ minutes.
# watchdog_thresh=140 and DEFAULT_HUNG_TASK_TIMEOUT=140 both result in detection after 140-280 secs.
#
# 4. Finally, "no output" crashes.
# sys/targets.Timeouts.NoOutput is set to 300 secs.
#
# Don't change these values without considering all other values as well.
# Note: executor also sets hung_task_check_interval_secs=20 sysctl.
# Note: gVisor watchdog timeout is 3 mins + 1/4 of that for checking period = 3m45s.
# Note: other arches that use qemu emulation scale all these timeouts by 3 (sys/targets.Timeouts.Scale).
- RCU_CPU_STALL_TIMEOUT: 100
# The constraints repeat constraints on inclusion of debug.yml which enables hung task detection.
- DEFAULT_HUNG_TASK_TIMEOUT: [140, -nonoise, -kmsan, -kcsan, -kmemleak]
# By default it is set to a too small value, which causes tons of
# `rcu_preempt detected expedited stalls on CPUs/tasks`.
# TODO: `rcu: Allow up to five minutes expedited RCU CPU stall-warning timeouts` lets us set more than 21k,
# but we first need to change the config generation tool to support multiple possible parameter values.
- RCU_EXP_CPU_STALL_TIMEOUT: [21000, v5.19]
|
