The interrupt is set using pseudo-NMI if it is available. Arm has a diagnostic interrupt feature called "Arm Generic Diagnostic Dump and Reset device", but the A64FX does not support this feature and instead has its own device definition. This patch series includes ones created by Sumit.[1] I tested on FX700: $ echo 1 > /proc/sys/kernel/sysrq $ echo HARDLOCKUP > /sys/kernel/debug/provoke-crash/DIRECT [ 124.712351] lkdtm: Performing direct entry HARDLOCKUP [ 147.232096] rcu: INFO: rcu_preempt detected stalls on CPUs/tasks: : : Send the "chassis power diag" command from the management server using ipmitool, the following message is shown: [ 206.061770] sysrq: Trigger a crash [ 206.061779] Kernel panic - not syncing: sysrq triggered crash : : Changes in V2: - Include Sumit's patch. - The handler calls handle_sysrq() to use the sysrq feature to cause a panic. - request_nmi() and request_irq() now use the same handler, and the function name of the handler has also changed. - Use readl()/writel() instead of readl_relaxed()/writel_relaxed(). V1: https://lore.kernel.org/linux-arm-kernel/20220218092010.1327309-1-hasegawa-hitomi@xxxxxxxxxxx/ [1] https://lore.kernel.org/all/20220228135408.1452882-1-sumit.garg@xxxxxxxxxx/ Sumit Garg (1): tty/sysrq: Make sysrq handler NMI aware Hitomi Hasegawa (1): soc: fujitsu: Add A64FX diagnostic interrupt driver MAINTAINERS | 5 + drivers/soc/Kconfig | 1 + drivers/soc/Makefile | 1 + drivers/soc/fujitsu/Kconfig | 13 +++ drivers/soc/fujitsu/Makefile | 3 + drivers/soc/fujitsu/a64fx-diag.c | 151 +++++++++++++++++++++++++++++++ drivers/tty/sysrq.c | 45 ++++++++- include/linux/sysrq.h | 1 + kernel/debug/debug_core.c | 1 + 9 files changed, 220 insertions(+), 1 deletion(-) create mode 100644 drivers/soc/fujitsu/Kconfig create mode 100644 drivers/soc/fujitsu/Makefile create mode 100644 drivers/soc/fujitsu/a64fx-diag.c -- 2.27.0
