[kvm-unit-tests GIT PULL 15/22] lib: s390x: terminate if PGM interrupt in interrupt handler

[Claudio Imbrenda <imbrenda@xxxxxxxxxxxxx>]

If a program interrupt is received while in an interrupt handler,
terminate immediately, stopping all CPUs and leaving the last CPU in
disabled wait with a specific PSW code.

This will aid debugging by not cluttering the output, avoiding further
interrupts (that would be needed to write to the output), and providing
an indication of the cause of the termination.

Signed-off-by: Claudio Imbrenda <imbrenda@xxxxxxxxxxxxx>
Message-Id: <20221020123143.213778-2-imbrenda@xxxxxxxxxxxxx>
Reviewed-by: Nico Boehr <nrb@xxxxxxxxxxxxx>
---
 lib/s390x/asm/arch_def.h | 11 +++++++++++
 lib/s390x/interrupt.c    | 20 ++++++++++++++++----
 2 files changed, 27 insertions(+), 4 deletions(-)

diff --git a/lib/s390x/asm/arch_def.h b/lib/s390x/asm/arch_def.h
index 41c9bd8c..783a7eaa 100644
--- a/lib/s390x/asm/arch_def.h
+++ b/lib/s390x/asm/arch_def.h
@@ -56,6 +56,7 @@ struct cpu {
 	bool active;
 	bool pgm_int_expected;
 	bool ext_int_expected;
+	bool in_interrupt_handler;
 };
 
 #define AS_PRIM				0
@@ -336,6 +337,16 @@ static inline void load_psw_mask(uint64_t mask)
 		: "+r" (tmp) :  "a" (&psw) : "memory", "cc" );
 }
 
+static inline void disabled_wait(uint64_t message)
+{
+	struct psw psw = {
+		.mask = PSW_MASK_WAIT,  /* Disabled wait */
+		.addr = message,
+	};
+
+	asm volatile("  lpswe 0(%0)\n" : : "a" (&psw) : "memory", "cc");
+}
+
 /**
  * psw_mask_clear_bits - clears bits from the current PSW mask
  * @clear: bitmask of bits that will be cleared
diff --git a/lib/s390x/interrupt.c b/lib/s390x/interrupt.c
index 7cc2c5fb..dadb7415 100644
--- a/lib/s390x/interrupt.c
+++ b/lib/s390x/interrupt.c
@@ -14,6 +14,7 @@
 #include <sie.h>
 #include <fault.h>
 #include <asm/page.h>
+#include "smp.h"
 
 /**
  * expect_pgm_int - Expect a program interrupt on the current CPU.
@@ -226,6 +227,11 @@ static void print_pgm_info(struct stack_frame_int *stack)
 
 void handle_pgm_int(struct stack_frame_int *stack)
 {
+	if (THIS_CPU->in_interrupt_handler) {
+		/* Something went very wrong, stop everything now without printing anything */
+		smp_teardown();
+		disabled_wait(0xfa12edbad21);
+	}
 	if (!THIS_CPU->pgm_int_expected) {
 		/* Force sclp_busy to false, otherwise we will loop forever */
 		sclp_handle_ext();
@@ -233,15 +239,18 @@ void handle_pgm_int(struct stack_frame_int *stack)
 	}
 
 	THIS_CPU->pgm_int_expected = false;
+	THIS_CPU->in_interrupt_handler = true;
 
 	if (THIS_CPU->pgm_cleanup_func)
 		THIS_CPU->pgm_cleanup_func(stack);
 	else
 		fixup_pgm_int(stack);
+	THIS_CPU->in_interrupt_handler = false;
 }
 
 void handle_ext_int(struct stack_frame_int *stack)
 {
+	THIS_CPU->in_interrupt_handler = true;
 	if (!THIS_CPU->ext_int_expected && lowcore.ext_int_code != EXT_IRQ_SERVICE_SIG) {
 		report_abort("Unexpected external call interrupt (code %#x): on cpu %d at %#lx",
 			     lowcore.ext_int_code, stap(), lowcore.ext_old_psw.addr);
@@ -260,6 +269,7 @@ void handle_ext_int(struct stack_frame_int *stack)
 
 	if (THIS_CPU->ext_cleanup_func)
 		THIS_CPU->ext_cleanup_func(stack);
+	THIS_CPU->in_interrupt_handler = false;
 }
 
 void handle_mcck_int(void)
@@ -272,11 +282,13 @@ static void (*io_int_func)(void);
 
 void handle_io_int(void)
 {
+	THIS_CPU->in_interrupt_handler = true;
 	if (io_int_func)
-		return io_int_func();
-
-	report_abort("Unexpected io interrupt: on cpu %d at %#lx",
-		     stap(), lowcore.io_old_psw.addr);
+		io_int_func();
+	else
+		report_abort("Unexpected io interrupt: on cpu %d at %#lx",
+			     stap(), lowcore.io_old_psw.addr);
+	THIS_CPU->in_interrupt_handler = false;
 }
 
 int register_io_int_func(void (*f)(void))
-- 
2.37.3