Merge tag 'arc-v3.13-rc1-part1' of git://git.kernel.org/pub/scm/linux…

…/kernel/git/vgupta/arc

Pull ARC changes from Vineet Gupta:
 - Towards a working SMP setup (ASID allocation, TLB Flush,...)
 - Support for TRACE_IRQFLAGS, LOCKDEP
 - cacheflush backend consolidation for I/D
 - Lots of allmodconfig fixlets from Chen
 - Other improvements/fixes

* tag 'arc-v3.13-rc1-part1' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc: (25 commits)
  ARC: [plat-arcfpga] defconfig update
  smp, ARC: kill SMP single function call interrupt
  ARC: [SMP] Disallow RTSC
  ARC: [SMP] Fix build failures for large NR_CPUS
  ARC: [SMP] enlarge possible NR_CPUS
  ARC: [SMP] TLB flush
  ARC: [SMP] ASID allocation
  arc: export symbol for pm_power_off in reset.c
  arc: export symbol for save_stack_trace() in stacktrace.c
  arc: remove '__init' for get_hw_config_num_irq()
  arc: remove '__init' for first_lines_of_secondary()
  arc: remove '__init' for setup_processor() and arc_init_IRQ()
  arc: kgdb: add default implementation for kgdb_roundup_cpus()
  ARC: Fix bogus gcc warning and micro-optimise TLB iteration loop
  ARC: Add support for irqflags tracing and lockdep
  ARC: Reset the value of Interrupt Priority Register
  ARC: Reduce #ifdef'ery for unaligned access emulation
  ARC: Change calling convention of do_page_fault()
  ARC: cacheflush optim - PTAG can be loop invariant if V-P is const
  ARC: cacheflush refactor #3: Unify the {d,i}cache flush leaf helpers
  ...

arch/arc/Kconfig

@@ -35,6 +35,12 @@ config ARC
 	select PERF_USE_VMALLOC
 	select HAVE_DEBUG_STACKOVERFLOW
 
+config TRACE_IRQFLAGS_SUPPORT
+	def_bool y
+
+config LOCKDEP_SUPPORT
+	def_bool y
+
 config SCHED_OMIT_FRAME_POINTER
 	def_bool y
 
@@ -130,17 +136,14 @@ if SMP
 config ARC_HAS_COH_CACHES
 	def_bool n
 
-config ARC_HAS_COH_RTSC
-	def_bool n
-
 config ARC_HAS_REENTRANT_IRQ_LV2
 	def_bool n
 
 endif
 
 config NR_CPUS
-	int "Maximum number of CPUs (2-32)"
-	range 2 32
+	int "Maximum number of CPUs (2-4096)"
+	range 2 4096
 	depends on SMP
 	default "2"
 
@@ -326,8 +329,7 @@ config ARC_HAS_RTSC
 	bool "Insn: RTSC (64-bit r/o cycle counter)"
 	default y
 	depends on ARC_CPU_REL_4_10
-	# if SMP, enable RTSC only if counter is coherent across cores
-	depends on !SMP || ARC_HAS_COH_RTSC
+	depends on !SMP
 
 endmenu   # "ARC CPU Configuration"
 

arch/arc/configs/fpga_defconfig

@@ -2,6 +2,8 @@ CONFIG_CROSS_COMPILE="arc-linux-uclibc-"
 # CONFIG_LOCALVERSION_AUTO is not set
 CONFIG_DEFAULT_HOSTNAME="ARCLinux"
 # CONFIG_SWAP is not set
+CONFIG_SYSVIPC=y
+CONFIG_POSIX_MQUEUE=y
 CONFIG_HIGH_RES_TIMERS=y
 CONFIG_IKCONFIG=y
 CONFIG_IKCONFIG_PROC=y
@@ -62,4 +64,5 @@ CONFIG_TMPFS=y
 CONFIG_NFS_FS=y
 # CONFIG_ENABLE_WARN_DEPRECATED is not set
 # CONFIG_ENABLE_MUST_CHECK is not set
+# CONFIG_DEBUG_PREEMPT is not set
 CONFIG_XZ_DEC=y

arch/arc/include/asm/cache.h

@@ -17,13 +17,7 @@
 #endif
 
 #define L1_CACHE_BYTES		(1 << L1_CACHE_SHIFT)
-
-/* For a rare case where customers have differently config I/D */
-#define ARC_ICACHE_LINE_LEN	L1_CACHE_BYTES
-#define ARC_DCACHE_LINE_LEN	L1_CACHE_BYTES
-
-#define ICACHE_LINE_MASK	(~(ARC_ICACHE_LINE_LEN - 1))
-#define DCACHE_LINE_MASK	(~(ARC_DCACHE_LINE_LEN - 1))
+#define CACHE_LINE_MASK		(~(L1_CACHE_BYTES - 1))
 
 /*
  * ARC700 doesn't cache any access in top 256M.

arch/arc/include/asm/irq.h

@@ -18,8 +18,8 @@
 
 #include <asm-generic/irq.h>
 
-extern void __init arc_init_IRQ(void);
-extern int __init get_hw_config_num_irq(void);
+extern void arc_init_IRQ(void);
+extern int get_hw_config_num_irq(void);
 
 void arc_local_timer_setup(unsigned int cpu);
 

arch/arc/include/asm/irqflags.h

@@ -151,16 +151,38 @@ static inline void arch_unmask_irq(unsigned int irq)
 
 #else
 
+#ifdef CONFIG_TRACE_IRQFLAGS
+
+.macro TRACE_ASM_IRQ_DISABLE
+	bl	trace_hardirqs_off
+.endm
+
+.macro TRACE_ASM_IRQ_ENABLE
+	bl	trace_hardirqs_on
+.endm
+
+#else
+
+.macro TRACE_ASM_IRQ_DISABLE
+.endm
+
+.macro TRACE_ASM_IRQ_ENABLE
+.endm
+
+#endif
+
 .macro IRQ_DISABLE  scratch
 	lr	\scratch, [status32]
 	bic	\scratch, \scratch, (STATUS_E1_MASK | STATUS_E2_MASK)
 	flag	\scratch
+	TRACE_ASM_IRQ_DISABLE
 .endm
 
 .macro IRQ_ENABLE  scratch
 	lr	\scratch, [status32]
 	or	\scratch, \scratch, (STATUS_E1_MASK | STATUS_E2_MASK)
 	flag	\scratch
+	TRACE_ASM_IRQ_ENABLE
 .endm
 
 #endif	/* __ASSEMBLY__ */

arch/arc/include/asm/mmu.h

@@ -48,7 +48,7 @@
 #ifndef __ASSEMBLY__
 
 typedef struct {
-	unsigned long asid;	/* 8 bit MMU PID + Generation cycle */
+	unsigned long asid[NR_CPUS];	/* 8 bit MMU PID + Generation cycle */
 } mm_context_t;
 
 #ifdef CONFIG_ARC_DBG_TLB_PARANOIA

arch/arc/include/asm/mmu_context.h

@@ -30,13 +30,13 @@
  * "Fast Context Switch" i.e. no TLB flush on ctxt-switch
  *
  * Linux assigns each task a unique ASID. A simple round-robin allocation
- * of H/w ASID is done using software tracker @asid_cache.
+ * of H/w ASID is done using software tracker @asid_cpu.
  * When it reaches max 255, the allocation cycle starts afresh by flushing
  * the entire TLB and wrapping ASID back to zero.
  *
  * A new allocation cycle, post rollover, could potentially reassign an ASID
  * to a different task. Thus the rule is to refresh the ASID in a new cycle.
- * The 32 bit @asid_cache (and mm->asid) have 8 bits MMU PID and rest 24 bits
+ * The 32 bit @asid_cpu (and mm->asid) have 8 bits MMU PID and rest 24 bits
  * serve as cycle/generation indicator and natural 32 bit unsigned math
  * automagically increments the generation when lower 8 bits rollover.
  */
@@ -47,16 +47,19 @@
 #define MM_CTXT_FIRST_CYCLE	(MM_CTXT_ASID_MASK + 1)
 #define MM_CTXT_NO_ASID		0UL
 
-#define hw_pid(mm)		(mm->context.asid & MM_CTXT_ASID_MASK)
+#define asid_mm(mm, cpu)	mm->context.asid[cpu]
+#define hw_pid(mm, cpu)		(asid_mm(mm, cpu) & MM_CTXT_ASID_MASK)
 
-extern unsigned int asid_cache;
+DECLARE_PER_CPU(unsigned int, asid_cache);
+#define asid_cpu(cpu)		per_cpu(asid_cache, cpu)
 
 /*
  * Get a new ASID if task doesn't have a valid one (unalloc or from prev cycle)
  * Also set the MMU PID register to existing/updated ASID
  */
 static inline void get_new_mmu_context(struct mm_struct *mm)
 {
+	const unsigned int cpu = smp_processor_id();
 	unsigned long flags;
 
 	local_irq_save(flags);
@@ -71,28 +74,28 @@ static inline void get_new_mmu_context(struct mm_struct *mm)
 	 * 	 first need to destroy the context, setting it to invalid
 	 * 	 value.
 	 */
-	if (!((mm->context.asid ^ asid_cache) & MM_CTXT_CYCLE_MASK))
+	if (!((asid_mm(mm, cpu) ^ asid_cpu(cpu)) & MM_CTXT_CYCLE_MASK))
 		goto set_hw;
 
 	/* move to new ASID and handle rollover */
-	if (unlikely(!(++asid_cache & MM_CTXT_ASID_MASK))) {
+	if (unlikely(!(++asid_cpu(cpu) & MM_CTXT_ASID_MASK))) {
 
-		flush_tlb_all();
+		local_flush_tlb_all();
 
 		/*
 		 * Above checke for rollover of 8 bit ASID in 32 bit container.
 		 * If the container itself wrapped around, set it to a non zero
 		 * "generation" to distinguish from no context
 		 */
-		if (!asid_cache)
-			asid_cache = MM_CTXT_FIRST_CYCLE;
+		if (!asid_cpu(cpu))
+			asid_cpu(cpu) = MM_CTXT_FIRST_CYCLE;
 	}
 
 	/* Assign new ASID to tsk */
-	mm->context.asid = asid_cache;
+	asid_mm(mm, cpu) = asid_cpu(cpu);
 
 set_hw:
-	write_aux_reg(ARC_REG_PID, hw_pid(mm) | MMU_ENABLE);
+	write_aux_reg(ARC_REG_PID, hw_pid(mm, cpu) | MMU_ENABLE);
 
 	local_irq_restore(flags);
 }
@@ -104,16 +107,45 @@ static inline void get_new_mmu_context(struct mm_struct *mm)
 static inline int
 init_new_context(struct task_struct *tsk, struct mm_struct *mm)
 {
-	mm->context.asid = MM_CTXT_NO_ASID;
+	int i;
+
+	for_each_possible_cpu(i)
+		asid_mm(mm, i) = MM_CTXT_NO_ASID;
+
 	return 0;
 }
 
+static inline void destroy_context(struct mm_struct *mm)
+{
+	unsigned long flags;
+
+	/* Needed to elide CONFIG_DEBUG_PREEMPT warning */
+	local_irq_save(flags);
+	asid_mm(mm, smp_processor_id()) = MM_CTXT_NO_ASID;
+	local_irq_restore(flags);
+}
+
 /* Prepare the MMU for task: setup PID reg with allocated ASID
     If task doesn't have an ASID (never alloc or stolen, get a new ASID)
 */
 static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 			     struct task_struct *tsk)
 {
+	const int cpu = smp_processor_id();
+
+	/*
+	 * Note that the mm_cpumask is "aggregating" only, we don't clear it
+	 * for the switched-out task, unlike some other arches.
+	 * It is used to enlist cpus for sending TLB flush IPIs and not sending
+	 * it to CPUs where a task once ran-on, could cause stale TLB entry
+	 * re-use, specially for a multi-threaded task.
+	 * e.g. T1 runs on C1, migrates to C3. T2 running on C2 munmaps.
+	 *      For a non-aggregating mm_cpumask, IPI not sent C1, and if T1
+	 *      were to re-migrate to C1, it could access the unmapped region
+	 *      via any existing stale TLB entries.
+	 */
+	cpumask_set_cpu(cpu, mm_cpumask(next));
+
 #ifndef CONFIG_SMP
 	/* PGD cached in MMU reg to avoid 3 mem lookups: task->mm->pgd */
 	write_aux_reg(ARC_REG_SCRATCH_DATA0, next->pgd);
@@ -131,11 +163,6 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
  */
 #define activate_mm(prev, next)		switch_mm(prev, next, NULL)
 
-static inline void destroy_context(struct mm_struct *mm)
-{
-	mm->context.asid = MM_CTXT_NO_ASID;
-}
-
 /* it seemed that deactivate_mm( ) is a reasonable place to do book-keeping
  * for retiring-mm. However destroy_context( ) still needs to do that because
  * between mm_release( ) = >deactive_mm( ) and

arch/arc/include/asm/setup.h

@@ -31,7 +31,7 @@ struct cpuinfo_data {
 extern int root_mountflags, end_mem;
 extern int running_on_hw;
 
-void __init setup_processor(void);
+void setup_processor(void);
 void __init setup_arch_memory(void);
 
 #endif /* __ASMARC_SETUP_H */

arch/arc/include/asm/smp.h

@@ -30,7 +30,7 @@ extern void arch_send_call_function_ipi_mask(const struct cpumask *mask);
  * APIs provided by arch SMP code to rest of arch code
  */
 extern void __init smp_init_cpus(void);
-extern void __init first_lines_of_secondary(void);
+extern void first_lines_of_secondary(void);
 extern const char *arc_platform_smp_cpuinfo(void);
 
 /*

arch/arc/include/asm/tlbflush.h

@@ -18,11 +18,18 @@ void local_flush_tlb_kernel_range(unsigned long start, unsigned long end);
 void local_flush_tlb_range(struct vm_area_struct *vma,
 			   unsigned long start, unsigned long end);
 
-/* XXX: Revisit for SMP */
+#ifndef CONFIG_SMP
 #define flush_tlb_range(vma, s, e)	local_flush_tlb_range(vma, s, e)
 #define flush_tlb_page(vma, page)	local_flush_tlb_page(vma, page)
 #define flush_tlb_kernel_range(s, e)	local_flush_tlb_kernel_range(s, e)
 #define flush_tlb_all()			local_flush_tlb_all()
 #define flush_tlb_mm(mm)		local_flush_tlb_mm(mm)
-
+#else
+extern void flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
+							 unsigned long end);
+extern void flush_tlb_page(struct vm_area_struct *vma, unsigned long page);
+extern void flush_tlb_kernel_range(unsigned long start, unsigned long end);
+extern void flush_tlb_all(void);
+extern void flush_tlb_mm(struct mm_struct *mm);
+#endif /* CONFIG_SMP */
 #endif

arch/arc/include/asm/unaligned.h

@@ -22,7 +22,8 @@ static inline int
 misaligned_fixup(unsigned long address, struct pt_regs *regs,
 		 struct callee_regs *cregs)
 {
-	return 0;
+	/* Not fixed */
+	return 1;
 }
 #endif
 

arch/arc/kernel/ctx_sw.c

@@ -17,6 +17,8 @@
 #include <asm/asm-offsets.h>
 #include <linux/sched.h>
 
+#define KSP_WORD_OFF 	((TASK_THREAD + THREAD_KSP) / 4)
+
 struct task_struct *__sched
 __switch_to(struct task_struct *prev_task, struct task_struct *next_task)
 {
@@ -45,7 +47,16 @@ __switch_to(struct task_struct *prev_task, struct task_struct *next_task)
 #endif
 
 		/* set ksp of outgoing task in tsk->thread.ksp */
+#if KSP_WORD_OFF <= 255
 		"st.as   sp, [%3, %1]    \n\t"
+#else
+		/*
+		 * Workaround for NR_CPUS=4k
+		 * %1 is bigger than 255 (S9 offset for st.as)
+		 */
+		"add2    r24, %3, %1     \n\t"
+		"st      sp, [r24]       \n\t"
+#endif
 
 		"sync   \n\t"
 
@@ -97,7 +108,7 @@ __switch_to(struct task_struct *prev_task, struct task_struct *next_task)
 		/* FP/BLINK restore generated by gcc (standard func epilogue */
 
 		: "=r"(tmp)
-		: "n"((TASK_THREAD + THREAD_KSP) / 4), "r"(next), "r"(prev)
+		: "n"(KSP_WORD_OFF), "r"(next), "r"(prev)
 		: "blink"
 	);
 

arch/arc/kernel/ctx_sw_asm.S

@@ -14,6 +14,8 @@
 #include <asm/asm-offsets.h>
 #include <asm/linkage.h>
 
+#define KSP_WORD_OFF 	((TASK_THREAD + THREAD_KSP) / 4)
+
 ;################### Low Level Context Switch ##########################
 
 	.section .sched.text,"ax",@progbits
@@ -28,8 +30,13 @@ __switch_to:
 	SAVE_CALLEE_SAVED_KERNEL
 
 	/* Save the now KSP in task->thread.ksp */
-	st.as  sp, [r0, (TASK_THREAD + THREAD_KSP)/4]
-
+#if KSP_WORD_OFF  <= 255
+	st.as  sp, [r0, KSP_WORD_OFF]
+#else
+	/* Workaround for NR_CPUS=4k as ST.as can only take s9 offset */
+	add2	r24, r0, KSP_WORD_OFF
+	st	sp, [r24]
+#endif
 	/*
 	* Return last task in r0 (return reg)
 	* On ARC, Return reg = First Arg reg = r0.