MIPS: Make irqflags.h functions preempt-safe for non-mipsr2 cpus

For non MIPSr2 processors, such as the BMIPS 5000, calls to
arch_local_irq_disable() and others may be preempted, and in doing
so a stale value may be restored to c0_status.  This fix disables
preemption for such processors prior to the call and enables it
after the call.

Those functions that needed this fix have been "outlined" to
mips-atomic.c, as they are no longer good candidates for inlining.

This bug was observed in a BMIPS 5000, occuring once every few hours
in a continuous reboot test.  It was traced to the write_lock_irq()
function which was being invoked in release_task() in exit.c.
By placing a number of "nops" inbetween the mfc0/mtc0 pair in
arch_local_irq_disable(), which is called by write_lock_irq(), we
were able to greatly increase the occurance of this bug.  Similarly,
the application of this commit silenced the bug.

Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
Cc: linux-mips@linux-mips.org
Cc: David Daney <ddaney.cavm@gmail.com>
Cc: Kevin Cernekee cernekee@gmail.com
Cc: Jim Quinlan <jim2101024@gmail.com>
Patchwork: https://patchwork.linux-mips.org/patch/4321/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
diff --git a/arch/mips/include/asm/irqflags.h b/arch/mips/include/asm/irqflags.h
index 309cbcd..9f3384c 100644
--- a/arch/mips/include/asm/irqflags.h
+++ b/arch/mips/include/asm/irqflags.h
@@ -16,6 +16,105 @@
 #include <linux/compiler.h>
 #include <asm/hazards.h>
 
+#if defined(CONFIG_CPU_MIPSR2) && !defined(CONFIG_MIPS_MT_SMTC)
+
+__asm__(
+	"	.macro	arch_local_irq_disable\n"
+	"	.set	push						\n"
+	"	.set	noat						\n"
+	"	di							\n"
+	"	irq_disable_hazard					\n"
+	"	.set	pop						\n"
+	"	.endm							\n");
+
+static inline void arch_local_irq_disable(void)
+{
+	__asm__ __volatile__(
+		"arch_local_irq_disable"
+		: /* no outputs */
+		: /* no inputs */
+		: "memory");
+}
+
+
+__asm__(
+	"	.macro	arch_local_irq_save result			\n"
+	"	.set	push						\n"
+	"	.set	reorder						\n"
+	"	.set	noat						\n"
+	"	di	\\result					\n"
+	"	andi	\\result, 1					\n"
+	"	irq_disable_hazard					\n"
+	"	.set	pop						\n"
+	"	.endm							\n");
+
+static inline unsigned long arch_local_irq_save(void)
+{
+	unsigned long flags;
+	asm volatile("arch_local_irq_save\t%0"
+		     : "=r" (flags)
+		     : /* no inputs */
+		     : "memory");
+	return flags;
+}
+
+
+__asm__(
+	"	.macro	arch_local_irq_restore flags			\n"
+	"	.set	push						\n"
+	"	.set	noreorder					\n"
+	"	.set	noat						\n"
+#if defined(CONFIG_IRQ_CPU)
+	/*
+	 * Slow, but doesn't suffer from a relatively unlikely race
+	 * condition we're having since days 1.
+	 */
+	"	beqz	\\flags, 1f					\n"
+	"	di							\n"
+	"	ei							\n"
+	"1:								\n"
+#else
+	/*
+	 * Fast, dangerous.  Life is fun, life is good.
+	 */
+	"	mfc0	$1, $12						\n"
+	"	ins	$1, \\flags, 0, 1				\n"
+	"	mtc0	$1, $12						\n"
+#endif
+	"	irq_disable_hazard					\n"
+	"	.set	pop						\n"
+	"	.endm							\n");
+
+static inline void arch_local_irq_restore(unsigned long flags)
+{
+	unsigned long __tmp1;
+
+	__asm__ __volatile__(
+		"arch_local_irq_restore\t%0"
+		: "=r" (__tmp1)
+		: "0" (flags)
+		: "memory");
+}
+
+static inline void __arch_local_irq_restore(unsigned long flags)
+{
+	unsigned long __tmp1;
+
+	__asm__ __volatile__(
+		"arch_local_irq_restore\t%0"
+		: "=r" (__tmp1)
+		: "0" (flags)
+		: "memory");
+}
+#else
+/* Functions that require preempt_{dis,en}able() are in mips-atomic.c */
+void arch_local_irq_disable(void);
+unsigned long arch_local_irq_save(void);
+void arch_local_irq_restore(unsigned long flags);
+void __arch_local_irq_restore(unsigned long flags);
+#endif /* if defined(CONFIG_CPU_MIPSR2) && !defined(CONFIG_MIPS_MT_SMTC) */
+
+
 __asm__(
 	"	.macro	arch_local_irq_enable				\n"
 	"	.set	push						\n"
@@ -57,55 +156,6 @@
 }
 
 
-/*
- * For cli() we have to insert nops to make sure that the new value
- * has actually arrived in the status register before the end of this
- * macro.
- * R4000/R4400 need three nops, the R4600 two nops and the R10000 needs
- * no nops at all.
- */
-/*
- * For TX49, operating only IE bit is not enough.
- *
- * If mfc0 $12 follows store and the mfc0 is last instruction of a
- * page and fetching the next instruction causes TLB miss, the result
- * of the mfc0 might wrongly contain EXL bit.
- *
- * ERT-TX49H2-027, ERT-TX49H3-012, ERT-TX49HL3-006, ERT-TX49H4-008
- *
- * Workaround: mask EXL bit of the result or place a nop before mfc0.
- */
-__asm__(
-	"	.macro	arch_local_irq_disable\n"
-	"	.set	push						\n"
-	"	.set	noat						\n"
-#ifdef CONFIG_MIPS_MT_SMTC
-	"	mfc0	$1, $2, 1					\n"
-	"	ori	$1, 0x400					\n"
-	"	.set	noreorder					\n"
-	"	mtc0	$1, $2, 1					\n"
-#elif defined(CONFIG_CPU_MIPSR2)
-	"	di							\n"
-#else
-	"	mfc0	$1,$12						\n"
-	"	ori	$1,0x1f						\n"
-	"	xori	$1,0x1f						\n"
-	"	.set	noreorder					\n"
-	"	mtc0	$1,$12						\n"
-#endif
-	"	irq_disable_hazard					\n"
-	"	.set	pop						\n"
-	"	.endm							\n");
-
-static inline void arch_local_irq_disable(void)
-{
-	__asm__ __volatile__(
-		"arch_local_irq_disable"
-		: /* no outputs */
-		: /* no inputs */
-		: "memory");
-}
-
 __asm__(
 	"	.macro	arch_local_save_flags flags			\n"
 	"	.set	push						\n"
@@ -125,113 +175,6 @@
 	return flags;
 }
 
-__asm__(
-	"	.macro	arch_local_irq_save result			\n"
-	"	.set	push						\n"
-	"	.set	reorder						\n"
-	"	.set	noat						\n"
-#ifdef CONFIG_MIPS_MT_SMTC
-	"	mfc0	\\result, $2, 1					\n"
-	"	ori	$1, \\result, 0x400				\n"
-	"	.set	noreorder					\n"
-	"	mtc0	$1, $2, 1					\n"
-	"	andi	\\result, \\result, 0x400			\n"
-#elif defined(CONFIG_CPU_MIPSR2)
-	"	di	\\result					\n"
-	"	andi	\\result, 1					\n"
-#else
-	"	mfc0	\\result, $12					\n"
-	"	ori	$1, \\result, 0x1f				\n"
-	"	xori	$1, 0x1f					\n"
-	"	.set	noreorder					\n"
-	"	mtc0	$1, $12						\n"
-#endif
-	"	irq_disable_hazard					\n"
-	"	.set	pop						\n"
-	"	.endm							\n");
-
-static inline unsigned long arch_local_irq_save(void)
-{
-	unsigned long flags;
-	asm volatile("arch_local_irq_save\t%0"
-		     : "=r" (flags)
-		     : /* no inputs */
-		     : "memory");
-	return flags;
-}
-
-__asm__(
-	"	.macro	arch_local_irq_restore flags			\n"
-	"	.set	push						\n"
-	"	.set	noreorder					\n"
-	"	.set	noat						\n"
-#ifdef CONFIG_MIPS_MT_SMTC
-	"mfc0	$1, $2, 1						\n"
-	"andi	\\flags, 0x400						\n"
-	"ori	$1, 0x400						\n"
-	"xori	$1, 0x400						\n"
-	"or	\\flags, $1						\n"
-	"mtc0	\\flags, $2, 1						\n"
-#elif defined(CONFIG_CPU_MIPSR2) && defined(CONFIG_IRQ_CPU)
-	/*
-	 * Slow, but doesn't suffer from a relatively unlikely race
-	 * condition we're having since days 1.
-	 */
-	"	beqz	\\flags, 1f					\n"
-	"	 di							\n"
-	"	ei							\n"
-	"1:								\n"
-#elif defined(CONFIG_CPU_MIPSR2)
-	/*
-	 * Fast, dangerous.  Life is fun, life is good.
-	 */
-	"	mfc0	$1, $12						\n"
-	"	ins	$1, \\flags, 0, 1				\n"
-	"	mtc0	$1, $12						\n"
-#else
-	"	mfc0	$1, $12						\n"
-	"	andi	\\flags, 1					\n"
-	"	ori	$1, 0x1f					\n"
-	"	xori	$1, 0x1f					\n"
-	"	or	\\flags, $1					\n"
-	"	mtc0	\\flags, $12					\n"
-#endif
-	"	irq_disable_hazard					\n"
-	"	.set	pop						\n"
-	"	.endm							\n");
-
-
-static inline void arch_local_irq_restore(unsigned long flags)
-{
-	unsigned long __tmp1;
-
-#ifdef CONFIG_MIPS_MT_SMTC
-	/*
-	 * SMTC kernel needs to do a software replay of queued
-	 * IPIs, at the cost of branch and call overhead on each
-	 * local_irq_restore()
-	 */
-	if (unlikely(!(flags & 0x0400)))
-		smtc_ipi_replay();
-#endif
-
-	__asm__ __volatile__(
-		"arch_local_irq_restore\t%0"
-		: "=r" (__tmp1)
-		: "0" (flags)
-		: "memory");
-}
-
-static inline void __arch_local_irq_restore(unsigned long flags)
-{
-	unsigned long __tmp1;
-
-	__asm__ __volatile__(
-		"arch_local_irq_restore\t%0"
-		: "=r" (__tmp1)
-		: "0" (flags)
-		: "memory");
-}
 
 static inline int arch_irqs_disabled_flags(unsigned long flags)
 {
@@ -245,7 +188,7 @@
 #endif
 }
 
-#endif
+#endif /* #ifndef __ASSEMBLY__ */
 
 /*
  * Do the CPU's IRQ-state tracing from assembly code.