5 files changed, 42 insertions, 22 deletions

diff --git a/arch/sparc/kernel/irq_64.c b/arch/sparc/kernel/irq_64.c
index 454ce3a2527..830d70a3e20 100644
--- a/arch/sparc/kernel/irq_64.c
+++ b/arch/sparc/kernel/irq_64.c
@@ -22,6 +22,7 @@
 #include <linux/seq_file.h>
 #include <linux/ftrace.h>
 #include <linux/irq.h>
+#include <linux/kmemleak.h>
 
 #include <asm/ptrace.h>
 #include <asm/processor.h>
@@ -46,6 +47,7 @@
 
 #include "entry.h"
 #include "cpumap.h"
+#include "kstack.h"
 
 #define NUM_IVECS       (IMAP_INR + 1)
 
@@ -712,24 +714,6 @@ void ack_bad_irq(unsigned int virt_irq)
 void *hardirq_stack[NR_CPUS];
 void *softirq_stack[NR_CPUS];
 
-static __attribute__((always_inline)) void *set_hardirq_stack(void)
-{
-        void *orig_sp, *sp = hardirq_stack[smp_processor_id()];
-
-        __asm__ __volatile__("mov %%sp, %0" : "=r" (orig_sp));
-        if (orig_sp < sp ||
-            orig_sp > (sp + THREAD_SIZE)) {
-                sp += THREAD_SIZE - 192 - STACK_BIAS;
-                __asm__ __volatile__("mov %0, %%sp" : : "r" (sp));
-        }
-
-        return orig_sp;
-}
-static __attribute__((always_inline)) void restore_hardirq_stack(void *orig_sp)
-{
-        __asm__ __volatile__("mov %0, %%sp" : : "r" (orig_sp));
-}
-
 void __irq_entry handler_irq(int irq, struct pt_regs *regs)
 {
         unsigned long pstate, bucket_pa;
diff --git a/arch/sparc/kernel/kstack.h b/arch/sparc/kernel/kstack.h
index 5247283d1c0..53dfb92e09f 100644
--- a/arch/sparc/kernel/kstack.h
+++ b/arch/sparc/kernel/kstack.h
@@ -61,4 +61,23 @@ check_magic:
 
 }
 
+static inline __attribute__((always_inline)) void *set_hardirq_stack(void)
+{
+        void *orig_sp, *sp = hardirq_stack[smp_processor_id()];
+
+        __asm__ __volatile__("mov %%sp, %0" : "=r" (orig_sp));
+        if (orig_sp < sp ||
+            orig_sp > (sp + THREAD_SIZE)) {
+                sp += THREAD_SIZE - 192 - STACK_BIAS;
+                __asm__ __volatile__("mov %0, %%sp" : : "r" (sp));
+        }
+
+        return orig_sp;
+}
+
+static inline __attribute__((always_inline)) void restore_hardirq_stack(void *orig_sp)
+{
+        __asm__ __volatile__("mov %0, %%sp" : : "r" (orig_sp));
+}
+
 #endif /* _KSTACK_H */
diff --git a/arch/sparc/kernel/nmi.c b/arch/sparc/kernel/nmi.c
index 75a3d1a2535..a4bd7ba74c8 100644
--- a/arch/sparc/kernel/nmi.c
+++ b/arch/sparc/kernel/nmi.c
@@ -23,6 +23,8 @@
 #include <asm/ptrace.h>
 #include <asm/pcr.h>
 
+#include "kstack.h"
+
 /* We don't have a real NMI on sparc64, but we can fake one
  * up using profiling counter overflow interrupts and interrupt
  * levels.
@@ -92,6 +94,7 @@ static void die_nmi(const char *str, struct pt_regs *regs, int do_panic)
 notrace __kprobes void perfctr_irq(int irq, struct pt_regs *regs)
 {
         unsigned int sum, touched = 0;
+        void *orig_sp;
 
         clear_softint(1 << irq);
 
@@ -99,6 +102,8 @@ notrace __kprobes void perfctr_irq(int irq, struct pt_regs *regs)
 
         nmi_enter();
 
+        orig_sp = set_hardirq_stack();
+
         if (notify_die(DIE_NMI, "nmi", regs, 0,
                        pt_regs_trap_type(regs), SIGINT) == NOTIFY_STOP)
                 touched = 1;
@@ -124,6 +129,8 @@ notrace __kprobes void perfctr_irq(int irq, struct pt_regs *regs)
                 pcr_ops->write(pcr_enable);
         }
 
+        restore_hardirq_stack(orig_sp);
+
         nmi_exit();
 }
 
diff --git a/arch/sparc/kernel/rtrap_64.S b/arch/sparc/kernel/rtrap_64.S
index 83f1873c6c1..090b9e9ad5e 100644
--- a/arch/sparc/kernel/rtrap_64.S
+++ b/arch/sparc/kernel/rtrap_64.S
@@ -130,7 +130,17 @@ rtrap_xcall:
                  nop
                 call                    trace_hardirqs_on
                  nop
-                wrpr                    %l4, %pil
+                /* Do not actually set the %pil here.  We will do that
+                 * below after we clear PSTATE_IE in the %pstate register.
+                 * If we re-enable interrupts here, we can recurse down
+                 * the hardirq stack potentially endlessly, causing a
+                 * stack overflow.
+                 *
+                 * It is tempting to put this test and trace_hardirqs_on
+                 * call at the 'rt_continue' label, but that will not work
+                 * as that path hits unconditionally and we do not want to
+                 * execute this in NMI return paths, for example.
+                 */
 #endif
 rtrap_no_irq_enable:
                 andcc                   %l1, TSTATE_PRIV, %l3
diff --git a/arch/sparc/kernel/unaligned_64.c b/arch/sparc/kernel/unaligned_64.c
index ebce43018c4..c752c4c479b 100644
--- a/arch/sparc/kernel/unaligned_64.c
+++ b/arch/sparc/kernel/unaligned_64.c
@@ -50,7 +50,7 @@ static inline enum direction decode_direction(unsigned int insn)
 }
 
 /* 16 = double-word, 8 = extra-word, 4 = word, 2 = half-word */
-static inline int decode_access_size(unsigned int insn)
+static inline int decode_access_size(struct pt_regs *regs, unsigned int insn)
 {
         unsigned int tmp;
 
@@ -66,7 +66,7 @@ static inline int decode_access_size(unsigned int insn)
                 return 2;
         else {
                 printk("Impossible unaligned trap. insn=%08x\n", insn);
-                die_if_kernel("Byte sized unaligned access?!?!", current_thread_info()->kregs);
+                die_if_kernel("Byte sized unaligned access?!?!", regs);
 
                 /* GCC should never warn that control reaches the end
                  * of this function without returning a value because
@@ -286,7 +286,7 @@ static void log_unaligned(struct pt_regs *regs)
 asmlinkage void kernel_unaligned_trap(struct pt_regs *regs, unsigned int insn)
 {
         enum direction dir = decode_direction(insn);
-        int size = decode_access_size(insn);
+        int size = decode_access_size(regs, insn);
         int orig_asi, asi;
 
         current_thread_info()->kern_una_regs = regs;