dashboard/config/linux: enable CSD_LOCK_WAIT_DEBUG

As asked by Hillf here:
https://lore.kernel.org/all/20220419134616.GE4285@paulmck-ThinkPad-P17-Gen-1/
This config can provide more precise diagnostic for some bugs.

5 files changed, 9 insertions, 4 deletions

diff --git a/dashboard/config/linux/bits/arm.yml b/dashboard/config/linux/bits/arm.yml
index b9b6337a6..a187b3ced 100644
--- a/dashboard/config/linux/bits/arm.yml
+++ b/dashboard/config/linux/bits/arm.yml
@@ -8,7 +8,7 @@ shell:
 config:
  # We reduce number of vivid/netrom/rose/hcd devices to 2 because we use at most 2 procs
  # with emulation and creating these devices takes time during boot.
- - CMDLINE: [append, "root=/dev/vda console=ttyAMA0 vmalloc=512M watchdog_thresh=165 workqueue.watchdog_thresh=420 sysctl.net.core.netdev_unregister_timeout_secs=420 dummy_hcd.num=2"]
+ - CMDLINE: [append, "root=/dev/vda console=ttyAMA0 vmalloc=512M smp.csd_lock_timeout=300000 watchdog_thresh=165 workqueue.watchdog_thresh=420 sysctl.net.core.netdev_unregister_timeout_secs=420 dummy_hcd.num=2"]
  - CMDLINE_EXTEND
  # See the comment in x86_64.yml re these numbers.
  - RCU_CPU_STALL_TIMEOUT: 300
diff --git a/dashboard/config/linux/bits/arm64.yml b/dashboard/config/linux/bits/arm64.yml
index 17d42f594..08ddfa04c 100644
--- a/dashboard/config/linux/bits/arm64.yml
+++ b/dashboard/config/linux/bits/arm64.yml
@@ -9,7 +9,7 @@ shell:
 config:
  # We reduce number of vivid/netrom/rose/hcd devices to 2 because we use at most 2 procs
  # with emulation and creating these devices takes time during boot.
- - CMDLINE: [append, "root=/dev/vda console=ttyAMA0 watchdog_thresh=165 workqueue.watchdog_thresh=420 sysctl.net.core.netdev_unregister_timeout_secs=420 dummy_hcd.num=2"]
+ - CMDLINE: [append, "root=/dev/vda console=ttyAMA0 smp.csd_lock_timeout=300000 watchdog_thresh=165 workqueue.watchdog_thresh=420 sysctl.net.core.netdev_unregister_timeout_secs=420 dummy_hcd.num=2"]
  # Was dropped in "arm64: Drop support for CMDLINE_EXTEND".
  - CMDLINE_EXTEND: [-v5.12]
  # See the comment in x86_64.yml re these numbers.
diff --git a/dashboard/config/linux/bits/debug.yml b/dashboard/config/linux/bits/debug.yml
index 099c02be2..aee3e760c 100644
--- a/dashboard/config/linux/bits/debug.yml
+++ b/dashboard/config/linux/bits/debug.yml
@@ -16,6 +16,10 @@ config:
  - DEBUG_VIRTUAL: [-s390]
  - LOCKUP_DETECTOR: [-s390]
  - SOFTLOCKUP_DETECTOR: [-s390]
+ # This config itself was added in v5.10, but the smp.csd_lock_timeout command line argument
+ # that is required to make it reliably usable is currently only in linux-next.
+ # TODO: change this to v5.18 when smp.csd_lock_timeout reaches mainline.
+ - CSD_LOCK_WAIT_DEBUG: [linux-next]
  - DETECT_HUNG_TASK
  - WQ_WATCHDOG
  - BOOTPARAM_HUNG_TASK_PANIC
diff --git a/dashboard/config/linux/bits/riscv64.yml b/dashboard/config/linux/bits/riscv64.yml
index 321bb2407..6501b02fa 100644
--- a/dashboard/config/linux/bits/riscv64.yml
+++ b/dashboard/config/linux/bits/riscv64.yml
@@ -6,7 +6,7 @@ shell:
  - make kvm_guest.config
 
 config:
- - CMDLINE: [append, "dummy_hcd.num=2 watchdog_thresh=165 workqueue.watchdog_thresh=420 sysctl.net.core.netdev_unregister_timeout_secs=420"]
+ - CMDLINE: [append, "dummy_hcd.num=2 smp.csd_lock_timeout=300000 watchdog_thresh=165 workqueue.watchdog_thresh=420 sysctl.net.core.netdev_unregister_timeout_secs=420"]
  - CMDLINE_EXTEND
  # See the comment in x86_64.yml re these numbers.
  - RCU_CPU_STALL_TIMEOUT: 300
diff --git a/dashboard/config/linux/bits/x86_64.yml b/dashboard/config/linux/bits/x86_64.yml
index b0391e5d8..583279143 100644
--- a/dashboard/config/linux/bits/x86_64.yml
+++ b/dashboard/config/linux/bits/x86_64.yml
@@ -9,7 +9,7 @@ config:
  # root=/dev/sda is not necessary correct, but it's a common setting and user arguments are appended
  # to the builtin command line, so user can always override this.
  # vsyscall=native required to run x86_64 executables on android kernels (it disables VDSO by default).
- - CMDLINE: [append, "root=/dev/sda console=ttyS0 vsyscall=native numa=fake=2 kvm-intel.nested=1 spec_store_bypass_disable=prctl nopcid vivid.n_devs=16 vivid.multiplanar=1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2 netrom.nr_ndevs=16 rose.rose_ndevs=16 dummy_hcd.num=8 watchdog_thresh=55 workqueue.watchdog_thresh=140 sysctl.net.core.netdev_unregister_timeout_secs=140"]
+ - CMDLINE: [append, "root=/dev/sda console=ttyS0 vsyscall=native numa=fake=2 kvm-intel.nested=1 spec_store_bypass_disable=prctl nopcid vivid.n_devs=16 vivid.multiplanar=1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2 netrom.nr_ndevs=16 rose.rose_ndevs=16 dummy_hcd.num=8 smp.csd_lock_timeout=100000 watchdog_thresh=55 workqueue.watchdog_thresh=140 sysctl.net.core.netdev_unregister_timeout_secs=140"]
  - CMDLINE_BOOL
 
  # Deterministic hang/stall detection.
@@ -20,6 +20,7 @@ config:
  #
  # 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.