As mentioned in bug 172213097 the buddy lockup detector does not provide meaningful debug information, i.e. stack dumps of the locked up core.

While a true hardlockup detector such as proposed in bug 172228850 would provide such information, not all platforms can support it. The platforms left out would still end up using the buddy lockup detector, or none at all.

Recently a series [1] that uses pseudo-NMIs on arm64 for IPIs allows getting NMI backtraces on hard locked up cores. This was tested on RK3399 on mainline [2] with:

echo HARDLOCKUP > /sys/kernel/debug/provoke-crash/DIRECT

Subsequently when an RCU stall is detected, an NMI backtrace is logged:

[   76.113761] lkdtm: Performing direct entry HARDLOCKUP
[  101.027946] rcu: INFO: rcu_sched detected stalls on CPUs/tasks:
[  101.028532] rcu: 	1-...0: (290 ticks this GP) idle=5c1/1/0x4000000000000000 softirq=1704/1705 fqs=2626 
[  101.029413] 	(detected by 0, t=5252 jiffies, g=1713, q=7)
[  101.029919] Sending NMI from CPU 0 to CPUs 1:
[  101.030350] NMI backtrace for cpu 1
[  101.030360] CPU: 1 PID: 1980 Comm: bash Not tainted 5.14.0-rc5-next-20210813-07973-g40f13b179b9c #5
[  101.030379] Hardware name: Firefly ROC-RK3399-PC Mezzanine Board (DT)
[  101.030388] pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[  101.030407] pc : lkdtm_HARDLOCKUP+0x24/0x70
[  101.030438] lr : lkdtm_do_action+0x24/0x30
[  101.030458] sp : ffffffc0116cbcf0
[  101.030464] pmr_save: 000000a0
[  101.030470] x29: ffffffc0116cbcf0 x28: ffffff8001b9a7c0 x27: 0000000000000000
[  101.030499] x26: 0000000000000000 x25: ffffffc010bcdb40 x24: ffffffc0116cbe00
[  101.030524] x23: 000000000000000b x22: ffffff8003ef7000 x21: ffffffc010d56b98
[  101.030548] x20: ffffffc010bcdd30 x19: 000000000000001f x18: 0000000000000000
[  101.030572] x17: 0000000000000000 x16: 0000000000000000 x15: 07500755074b0743
[  101.030596] x14: 074f074c07440752 x13: ffffffc011090bd0 x12: 0000000000000570
[  101.030621] x11: 00000000000001d0 x10: ffffffc0110e8bd0 x9 : 00000000fffff000
[  101.030645] x8 : ffffffc011090bd0 x7 : ffffffc0110e8bd0 x6 : 0000000000000001
[  101.030668] x5 : 0000000000000000 x4 : ffffff80f777ca68 x3 : 0000000000000000
[  101.030691] x2 : 0000000000000000 x1 : ffffff8001b9a7c0 x0 : 00000000000000a0
[  101.030715] Call trace:
[  101.030723]  lkdtm_HARDLOCKUP+0x24/0x70
[  101.030746]  direct_entry+0x124/0x1bc
[  101.030767]  full_proxy_write+0x64/0xc0
[  101.030788]  vfs_write+0xf4/0x2c4
[  101.030811]  ksys_write+0x68/0xf4
[  101.030831]  __arm64_sys_write+0x20/0x2c
[  101.030853]  invoke_syscall+0x48/0x114
[  101.030875]  el0_svc_common.constprop.0+0x4c/0x15c
[  101.030895]  do_el0_svc+0x28/0x90
[  101.030914]  el0_svc+0x18/0x54
[  101.030929]  el0t_64_sync_handler+0x1a8/0x1b0
[  101.030947]  el0t_64_sync+0x1a0/0x1a4

SysRq backtraces also work properly.

Without pseudo-NMIs enabled, only a task dump is logged:

[  228.387245] rcu: INFO: rcu_sched detected stalls on CPUs/tasks:
[  228.387835] rcu: 	1-...0: (39 ticks this GP) idle=b39/1/0x4000000000000000 softirq=1451/1452 fqs=10492 
[  228.388716] 	(detected by 0, t=21007 jiffies, g=1517, q=321)
[  228.389244] Task dump for CPU 1:
[  228.389551] task:bash            state:R  running task     stack:    0 pid: 1955 ppid:  1954 flags:0x00000002
[  228.390477] Call trace:
[  228.390711]  __switch_to+0x104/0x174
[  228.391073]  0xffffff800114b500

And SysRq backtraces aren't available for all cores.

[1] https://lore.kernel.org/linux-arm-kernel/1604317487-14543-1-git-send-email-sumit.garg@linaro.org/
[2] https://lore.kernel.org/linux-arm-kernel/CAGb2v66mVoWiCibjq25d3Z8OvbWNO9p+vMo761RJLiD-BqVbqw@mail.gmail.com/