[PATCH] x86/platform/uv: Add more to slave cpu kdump info

From: Mike Travis
Date: Thu Jan 21 2021 - 09:36:15 EST


Add call to run_crash_ipi_callback() to gather more info of what the
slave cpus were doing to help with failure analysis.

Excerpt from Georges:
'It is only changing where crash slaves will be stalling after having
taken care of properly laying down "crash note regs". Please note that
"crash note regs" are a key piece of data used by crash dump debuggers
to provide a reliable backtrace of running processors.'

Signed-off-by: Mike Travis <mike.travis@xxxxxxx>
Signed-off-by: Georges Aureau <georges.aureau@xxxxxxx>
Reviewed-by: Steve Wahl <steve.wahl@xxxxxxx>
---
arch/x86/platform/uv/uv_nmi.c | 30 +++++++++++++++++++-----------
1 file changed, 19 insertions(+), 11 deletions(-)

diff --git a/arch/x86/platform/uv/uv_nmi.c b/arch/x86/platform/uv/uv_nmi.c
index eafc530c8767..811997b16b42 100644
--- a/arch/x86/platform/uv/uv_nmi.c
+++ b/arch/x86/platform/uv/uv_nmi.c
@@ -24,6 +24,7 @@
#include <asm/kdebug.h>
#include <asm/local64.h>
#include <asm/nmi.h>
+#include <asm/reboot.h>
#include <asm/traps.h>
#include <asm/uv/uv.h>
#include <asm/uv/uv_hub.h>
@@ -834,28 +835,35 @@ static void uv_nmi_touch_watchdogs(void)
touch_nmi_watchdog();
}

-static atomic_t uv_nmi_kexec_failed;
-
#if defined(CONFIG_KEXEC_CORE)
+static atomic_t uv_nmi_kexec_failed;
static void uv_nmi_kdump(int cpu, int master, struct pt_regs *regs)
{
+ /* Check if kdump kernel loaded for both master and slave CPUs */
+ if (!kexec_crash_image) {
+ if (master)
+ pr_err("UV: NMI error: kdump kernel not loaded\n");
+ return;
+ }
+
/* Call crash to dump system state */
if (master) {
pr_emerg("UV: NMI executing crash_kexec on CPU%d\n", cpu);
crash_kexec(regs);

- pr_emerg("UV: crash_kexec unexpectedly returned, ");
+ pr_emerg("UV: crash_kexec unexpectedly returned\n");
atomic_set(&uv_nmi_kexec_failed, 1);
- if (!kexec_crash_image) {
- pr_cont("crash kernel not loaded\n");
- return;
+ } else {
+ /* If kdump kernel fails, slaves will exit this loop */
+ while (atomic_read(&uv_nmi_kexec_failed) == 0) {
+ /*
+ * If crash exec sets "crash_ipi_issued",
+ * ... slaves do not return from callback function.
+ */
+ run_crash_ipi_callback(regs);
+ mdelay(10);
}
- pr_cont("kexec busy, stalling cpus while waiting\n");
}
-
- /* If crash exec fails the slaves should return, otherwise stall */
- while (atomic_read(&uv_nmi_kexec_failed) == 0)
- mdelay(10);
}

#else /* !CONFIG_KEXEC_CORE */
--
2.21.0