linux-toradex.git/kernel/watchdog.c, branch v7.0-rc7

watchdog/softlockup: fix sample ring index wrap in need_counting_irqs()

2026-02-08T08:13:34+00:00

cpustat_tail indexes cpustat_util[], which is a NUM_SAMPLE_PERIODS-sized
ring buffer. need_counting_irqs() currently wraps the index using
NUM_HARDIRQ_REPORT, which only happens to match NUM_SAMPLE_PERIODS.

Use NUM_SAMPLE_PERIODS for the wrap to keep the ring math correct even if
the NUM_HARDIRQ_REPORT or  NUM_SAMPLE_PERIODS changes.

Link: https://lkml.kernel.org/r/tencent_7068189CB6D6689EB353F3D17BF5A5311A07@qq.com
Fixes: e9a9292e2368 ("watchdog/softlockup: Report the most frequent interrupts")
Signed-off-by: Shengming Hu 
Reviewed-by: Petr Mladek 
Cc: Ingo Molnar 
Cc: Mark Brown 
Cc: Thomas Gleixner 
Cc: Zhang Run 
Cc: 
Signed-off-by: Andrew Morton

watchdog: softlockup: panic when lockup duration exceeds N thresholds

2026-01-21T03:44:20+00:00

The softlockup_panic sysctl is currently a binary option: panic
immediately or never panic on soft lockups.

Panicking on any soft lockup, regardless of duration, can be overly
aggressive for brief stalls that may be caused by legitimate operations. 
Conversely, never panicking may allow severe system hangs to persist
undetected.

Extend softlockup_panic to accept an integer threshold, allowing the
kernel to panic only when the normalized lockup duration exceeds N
watchdog threshold periods.  This provides finer-grained control to
distinguish between transient delays and persistent system failures.

The accepted values are:
- 0: Don't panic (unchanged)
- 1: Panic when duration >= 1 * threshold (20s default, original behavior)
- N > 1: Panic when duration >= N * threshold (e.g., 2 = 40s, 3 = 60s.)

The original behavior is preserved for values 0 and 1, maintaining full
backward compatibility while allowing systems to tolerate brief lockups
while still catching severe, persistent hangs.

[lirongqing@baidu.com: v2]
  Link: https://lkml.kernel.org/r/20251218074300.4080-1-lirongqing@baidu.com
Link: https://lkml.kernel.org/r/20251216074521.2796-1-lirongqing@baidu.com
Signed-off-by: Li RongQing 
Cc: Eduard Zingerman 
Cc: Hao Luo 
Cc: Jiri Olsa 
Cc: John Fastabend 
Cc: KP Singh 
Cc: Lance Yang 
Cc: Martin KaFai Lau 
Cc: Nicholas Piggin 
Cc: Song Liu 
Cc: Stanislav Fomichev 
Cc: Yonghong Song 
Signed-off-by: Andrew Morton

powerpc/watchdog: add support for hardlockup_sys_info sysctl

2026-01-15T06:16:22+00:00

Commit a9af76a78760 ("watchdog: add sys_info sysctls to dump sys info on
system lockup") adds 'hardlock_sys_info' systcl knob for general kernel
watchdog to control what kinds of system debug info to be dumped on
hardlockup.

Add similar support in powerpc watchdog code to make the sysctl knob more
general, which also fixes a compiling warning in general watchdog code
reported by 0day bot.

Link: https://lkml.kernel.org/r/20251231080309.39642-1-feng.tang@linux.alibaba.com
Fixes: a9af76a78760 ("watchdog: add sys_info sysctls to dump sys info on system lockup")
Signed-off-by: Feng Tang 
Reported-by: kernel test robot 
Closes: https://lore.kernel.org/oe-kbuild-all/202512030920.NFKtekA7-lkp@intel.com/
Suggested-by: Petr Mladek 
Reviewed-by: Petr Mladek 
Cc: Madhavan Srinivasan 
Cc: Michael Ellerman 
Cc: Nicholas Piggin 
Signed-off-by: Andrew Morton

Merge tag 'mm-nonmm-stable-2025-12-06-11-14' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm

2025-12-06T22:01:20+00:00

Pull non-MM updates from Andrew Morton:

 - "panic: sys_info: Refactor and fix a potential issue" (Andy Shevchenko)
   fixes a build issue and does some cleanup in ib/sys_info.c

 - "Implement mul_u64_u64_div_u64_roundup()" (David Laight)
   enhances the 64-bit math code on behalf of a PWM driver and beefs up
   the test module for these library functions

 - "scripts/gdb/symbols: make BPF debug info available to GDB" (Ilya Leoshkevich)
   makes BPF symbol names, sizes, and line numbers available to the GDB
   debugger

 - "Enable hung_task and lockup cases to dump system info on demand" (Feng Tang)
   adds a sysctl which can be used to cause additional info dumping when
   the hung-task and lockup detectors fire

 - "lib/base64: add generic encoder/decoder, migrate users" (Kuan-Wei Chiu)
   adds a general base64 encoder/decoder to lib/ and migrates several
   users away from their private implementations

 - "rbree: inline rb_first() and rb_last()" (Eric Dumazet)
   makes TCP a little faster

 - "liveupdate: Rework KHO for in-kernel users" (Pasha Tatashin)
   reworks the KEXEC Handover interfaces in preparation for Live Update
   Orchestrator (LUO), and possibly for other future clients

 - "kho: simplify state machine and enable dynamic updates" (Pasha Tatashin)
   increases the flexibility of KEXEC Handover. Also preparation for LUO

 - "Live Update Orchestrator" (Pasha Tatashin)
   is a major new feature targeted at cloud environments. Quoting the
   cover letter:

      This series introduces the Live Update Orchestrator, a kernel
      subsystem designed to facilitate live kernel updates using a
      kexec-based reboot. This capability is critical for cloud
      environments, allowing hypervisors to be updated with minimal
      downtime for running virtual machines. LUO achieves this by
      preserving the state of selected resources, such as memory,
      devices and their dependencies, across the kernel transition.

      As a key feature, this series includes support for preserving
      memfd file descriptors, which allows critical in-memory data, such
      as guest RAM or any other large memory region, to be maintained in
      RAM across the kexec reboot.

   Mike Rappaport merits a mention here, for his extensive review and
   testing work.

 - "kexec: reorganize kexec and kdump sysfs" (Sourabh Jain)
   moves the kexec and kdump sysfs entries from /sys/kernel/ to
   /sys/kernel/kexec/ and adds back-compatibility symlinks which can
   hopefully be removed one day

 - "kho: fixes for vmalloc restoration" (Mike Rapoport)
   fixes a BUG which was being hit during KHO restoration of vmalloc()
   regions

* tag 'mm-nonmm-stable-2025-12-06-11-14' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (139 commits)
  calibrate: update header inclusion
  Reinstate "resource: avoid unnecessary lookups in find_next_iomem_res()"
  vmcoreinfo: track and log recoverable hardware errors
  kho: fix restoring of contiguous ranges of order-0 pages
  kho: kho_restore_vmalloc: fix initialization of pages array
  MAINTAINERS: TPM DEVICE DRIVER: update the W-tag
  init: replace simple_strtoul with kstrtoul to improve lpj_setup
  KHO: fix boot failure due to kmemleak access to non-PRESENT pages
  Documentation/ABI: new kexec and kdump sysfs interface
  Documentation/ABI: mark old kexec sysfs deprecated
  kexec: move sysfs entries to /sys/kernel/kexec
  test_kho: always print restore status
  kho: free chunks using free_page() instead of kfree()
  selftests/liveupdate: add kexec test for multiple and empty sessions
  selftests/liveupdate: add simple kexec-based selftest for LUO
  selftests/liveupdate: add userspace API selftests
  docs: add documentation for memfd preservation via LUO
  mm: memfd_luo: allow preserving memfd
  liveupdate: luo_file: add private argument to store runtime state
  mm: shmem: export some functions to internal.h
  ...

Merge tag 'sysctl-6.19-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/sysctl/sysctl

2025-12-05T19:15:37+00:00

Pull sysctl updates from Joel Granados:

 - Move jiffies converters out of kernel/sysctl.c

   Move the jiffies converters into kernel/time/jiffies.c and replace
   the pipe-max-size proc_handler converter with a macro based version.
   This is all part of the effort to relocate non-sysctl logic out of
   kernel/sysctl.c into more relevant subsystems. No functional changes.

 - Generalize proc handler converter creation

   Remove duplicated sysctl converter logic by consolidating it in
   macros. These are used inside sysctl core as well as in pipe.c and
   jiffies.c. Converter kernel and user space pointer args are now
   automatically const qualified for the convenience of the caller. No
   functional changes.

 - Miscellaneous

   Fix kernel-doc format warnings, remove unnecessary __user
   qualifiers, and move the nmi_watchdog sysctl into .rodata.

 - Testing

   This series was run through sysctl selftests/kunit test suite in
   x86_64. It went into linux-next after rc2, giving it a good 4/5 weeks
   of testing.

* tag 'sysctl-6.19-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/sysctl/sysctl: (21 commits)
  sysctl: Wrap do_proc_douintvec with the public function proc_douintvec_conv
  sysctl: Create pipe-max-size converter using sysctl UINT macros
  sysctl: Move proc_doulongvec_ms_jiffies_minmax to kernel/time/jiffies.c
  sysctl: Move jiffies converters to kernel/time/jiffies.c
  sysctl: Move UINT converter macros to sysctl header
  sysctl: Move INT converter macros to sysctl header
  sysctl: Allow custom converters from outside sysctl
  sysctl: remove __user qualifier from stack_erasing_sysctl buffer argument
  sysctl: Create macro for user-to-kernel uint converter
  sysctl: Add optional range checking to SYSCTL_UINT_CONV_CUSTOM
  sysctl: Create unsigned int converter using new macro
  sysctl: Add optional range checking to SYSCTL_INT_CONV_CUSTOM
  sysctl: Create integer converters with one macro
  sysctl: Create converter functions with two new macros
  sysctl: Discriminate between kernel and user converter params
  sysctl: Indicate the direction of operation with macro names
  sysctl: Remove superfluous __do_proc_* indirection
  sysctl: Remove superfluous tbl_data param from "dovec" functions
  sysctl: Replace void pointer with const pointer to ctl_table
  sysctl: fix kernel-doc format warning
  ...

watchdog: add sys_info sysctls to dump sys info on system lockup

2025-11-20T22:03:43+00:00

When soft/hard lockup happens, developers may need different kinds of
system information (call-stacks, memory info, locks, etc.) to help
debugging.

Add 'softlockup_sys_info' and 'hardlockup_sys_info' sysctl knobs to take
human readable string like "tasks,mem,timers,locks,ftrace,...", and when
system lockup happens, all requested information will be printed out. 
(refer kernel/sys_info.c for more details).

Link: https://lkml.kernel.org/r/20251113111039.22701-4-feng.tang@linux.alibaba.com
Signed-off-by: Feng Tang 
Reviewed-by: Petr Mladek 
Cc: Jonathan Corbet 
Cc: Lance Yang 
Cc: "Paul E . McKenney" 
Cc: Petr Mladek 
Cc: Steven Rostedt 
Signed-off-by: Andrew Morton

sched_ext: Pass locked CPU parameter to scx_hardlockup() and add docs

2025-11-14T21:11:08+00:00

With the buddy lockup detector, smp_processor_id() returns the detecting CPU,
not the locked CPU, making scx_hardlockup()'s printouts confusing. Pass the
locked CPU number from watchdog_hardlockup_check() as a parameter instead.

Also add kerneldoc comments to handle_lockup(), scx_hardlockup(), and
scx_rcu_cpu_stall() documenting their return value semantics.

Suggested-by: Doug Anderson 
Reviewed-by: Douglas Anderson 
Acked-by: Andrea Righi 
Reviewed-by: Emil Tsalapatis 
Signed-off-by: Tejun Heo

sched_ext: Hook up hardlockup detector

2025-11-12T16:43:44+00:00

A poorly behaving BPF scheduler can trigger hard lockup. For example, on a
large system with many tasks pinned to different subsets of CPUs, if the BPF
scheduler puts all tasks in a single DSQ and lets all CPUs at it, the DSQ lock
can be contended to the point where hardlockup triggers. Unfortunately,
hardlockup can be the first signal out of such situations, thus requiring
hardlockup handling.

Hook scx_hardlockup() into the hardlockup detector to try kicking out the
current scheduler in an attempt to recover the system to a good state. The
handling strategy can delay watchdog taking its own action by one polling
period; however, given that the only remediation for hardlockup is crash, this
is likely an acceptable trade-off.

v2: Add missing dummy scx_hardlockup() definition for
    !CONFIG_SCHED_CLASS_EXT (kernel test bot).

Reported-by: Dan Schatzberg 
Cc: Emil Tsalapatis 
Cc: Douglas Anderson 
Cc: Andrew Morton 
Reviewed-by: Andrea Righi 
Signed-off-by: Tejun Heo

watchdog: move nmi_watchdog sysctl into .rodata

2025-10-24T13:35:11+00:00

Move nmi_watchdog into the watchdog_sysctls array to prevent it from
unnecessary modification. This move effectively moves it inside the
.rodata section.

Initially moved out into its own non-const array in commit 9ec272c586b0
("watchdog/hardlockup: keep kernel.nmi_watchdog sysctl as 0444 if probe
fails"), which made it writable only when watchdog_hardlockup_available
was true. Moving it back to watchdog_sysctl keeps this behavior as
writing to nmi_watchdog still fails when watchdog_hardlockup_available
is false.

Reviewed-by: Petr Mladek 
Signed-off-by: Joel Granados

watchdog: skip checks when panic is in progress

2025-09-14T00:32:53+00:00

This issue was found when an EFI pstore was configured for kdump logging
with the NMI hard lockup detector enabled.  The efi-pstore write operation
was slow, and with a large number of logs, the pstore dump callback within
kmsg_dump() took a long time.

This delay triggered the NMI watchdog, leading to a nested panic.  The
call flow demonstrates how the secondary panic caused an
emergency_restart() to be triggered before the initial pstore operation
could finish, leading to a failure to dump the logs:

  real panic() {
	kmsg_dump() {
		...
		pstore_dump() {
			start_dump();
			... // long time operation triggers NMI watchdog
			nmi panic() {
				...
				emergency_restart(); // pstore unfinished
			}
			...
			finish_dump(); // never reached
		}
	}
  }

Both watchdog_buddy_check_hardlockup() and watchdog_overflow_callback()
may trigger during a panic.  This can lead to recursive panic handling.

Add panic_in_progress() checks so watchdog activity is skipped once a
panic has begun.

This prevents recursive panic and keeps the panic path more reliable.

Link: https://lkml.kernel.org/r/20250825022947.1596226-10-wangjinchao600@gmail.com
Signed-off-by: Jinchao Wang 
Reviewed-by: Yury Norov (NVIDIA) 
Cc: Anna Schumaker 
Cc: Baoquan He 
Cc: "Darrick J. Wong" 
Cc: Dave Young 
Cc: Doug Anderson 
Cc: "Guilherme G. Piccoli" 
Cc: Helge Deller 
Cc: Ingo Molnar 
Cc: Jason Gunthorpe 
Cc: Joanthan Cameron 
Cc: Joel Granados 
Cc: John Ogness 
Cc: Kees Cook 
Cc: Li Huafei 
Cc: "Luck, Tony" 
Cc: Luo Gengkun 
Cc: Max Kellermann 
Cc: Nam Cao 
Cc: oushixiong 
Cc: Petr Mladek 
Cc: Qianqiang Liu 
Cc: Sergey Senozhatsky 
Cc: Sohil Mehta 
Cc: Steven Rostedt 
Cc: Tejun Heo 
Cc: Thomas Gleinxer 
Cc: Thomas Zimemrmann 
Cc: Thorsten Blum 
Cc: Ville Syrjala 
Cc: Vivek Goyal 
Cc: Yicong Yang 
Cc: Yunhui Cui 
Signed-off-by: Andrew Morton