[Skiboot] [RFC PATCH 3/3] core/vm: try to handle recoverable MCEs by turning off VMM

[Nicholas Piggin]

This is not foolproof because local mappings and io accessors will
not cope with the VMM being switched off, but it may catch some
cases. io accessors in particular won't recover properly because
they won't switch to use the real-mode cache inhibited instructions
after VMM turns off. That could be solved by marking IO accessors
as non-recoverable, or providing an exception recovery address
which retries.

This is mostly just a debugging aid, real MCE recovery for VMM
faults would have to flush SLB and translation caches.
---
 core/exceptions.c | 26 ++++++++++++++++++++++++++
 1 file changed, 26 insertions(+)

diff --git a/core/exceptions.c b/core/exceptions.c
index 89b4451ab..8dafed676 100644
--- a/core/exceptions.c
+++ b/core/exceptions.c
@@ -41,6 +41,21 @@ static void dump_regs(struct stack_frame *stack)
 
 #define EXCEPTION_MAX_STR 320
 
+static void print_recoverable_mce_vm(struct stack_frame *stack, uint64_t nip, uint64_t msr)
+{
+	char buf[EXCEPTION_MAX_STR];
+	size_t l;
+
+	l = 0;
+	l += snprintf(buf + l, EXCEPTION_MAX_STR - l,
+		"Recoverable MCE with VM on at "REG"   ", nip);
+	l += snprintf_symbol(buf + l, EXCEPTION_MAX_STR - l, nip);
+	l += snprintf(buf + l, EXCEPTION_MAX_STR - l, "  MSR "REG, msr);
+	prerror("%s\n", buf);
+	dump_regs(stack);
+	prerror("Continuing with VM off\n");
+}
+
 void exception_entry(struct stack_frame *stack)
 {
 	bool fatal = false;
@@ -92,6 +107,17 @@ void exception_entry(struct stack_frame *stack)
 		break;
 
 	case 0x200:
+		if (this_cpu()->vm_local_map_inuse)
+			fatal = true; /* local map is non-linear */
+
+		if (!fatal && (msr & (MSR_IR|MSR_DR))) {
+			print_recoverable_mce_vm(stack, nip, msr);
+			/* Turn off VM and try again */
+			this_cpu()->vm_setup = false;
+			stack->srr1 &= ~(MSR_IR|MSR_DR);
+			goto out;
+		}
+
 		fatal = true;
 		prerror("***********************************************\n");
 		l += snprintf(buf + l, EXCEPTION_MAX_STR - l,
-- 
2.20.1