powerpc/64s/radix: refactor TLB flush type selection

The logic to decide what kind of TLB flush is required (local, global,
or IPI) is spread multiple times over the several kinds of TLB flushes.

Move it all into a single function which may issue IPIs if necessary,
and also returns a flush type that is to be used.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20201217134731.488135-3-npiggin@gmail.com

arch/powerpc/mm/book3s64/radix_tlb.c

@@ -627,15 +627,6 @@ void radix__local_flush_tlb_page(struct vm_area_struct *vma, unsigned long vmadd
 }
 EXPORT_SYMBOL(radix__local_flush_tlb_page);
 
-static bool mm_is_singlethreaded(struct mm_struct *mm)
-{
-	if (atomic_read(&mm->context.copros) > 0)
-		return false;
-	if (atomic_read(&mm->mm_users) <= 1 && current->mm == mm)
-		return true;
-	return false;
-}
-
 static bool mm_needs_flush_escalation(struct mm_struct *mm)
 {
 	/*
@@ -707,10 +698,58 @@ static void exit_flush_lazy_tlbs(struct mm_struct *mm)
 	smp_call_function_many(mm_cpumask(mm), do_exit_flush_lazy_tlb,
 				(void *)mm, 1);
 }
+#else /* CONFIG_SMP */
+static inline void exit_flush_lazy_tlbs(struct mm_struct *mm) { }
+#endif /* CONFIG_SMP */
+
+enum tlb_flush_type {
+	FLUSH_TYPE_LOCAL,
+	FLUSH_TYPE_GLOBAL,
+};
+
+static enum tlb_flush_type flush_type_needed(struct mm_struct *mm, bool fullmm)
+{
+	if (mm_is_thread_local(mm))
+		return FLUSH_TYPE_LOCAL;
+
+	/* Coprocessors require TLBIE to invalidate nMMU. */
+	if (atomic_read(&mm->context.copros) > 0)
+		return FLUSH_TYPE_GLOBAL;
+
+	/*
+	 * In the fullmm case there's no point doing the exit_flush_lazy_tlbs
+	 * because the mm is being taken down anyway, and a TLBIE tends to
+	 * be faster than an IPI+TLBIEL.
+	 */
+	if (fullmm)
+		return FLUSH_TYPE_GLOBAL;
+
+	/*
+	 * If we are running the only thread of a single-threaded process,
+	 * then we should almost always be able to trim off the rest of the
+	 * CPU mask (except in the case of use_mm() races), so always try
+	 * trimming the mask.
+	 */
+	if (atomic_read(&mm->mm_users) <= 1 && current->mm == mm) {
+		exit_flush_lazy_tlbs(mm);
+		/*
+		 * use_mm() race could prevent IPIs from being able to clear
+		 * the cpumask here, however those users are established
+		 * after our first check (and so after the PTEs are removed),
+		 * and the TLB still gets flushed by the IPI, so this CPU
+		 * will only require a local flush.
+		 */
+		return FLUSH_TYPE_LOCAL;
+	}
+
+	return FLUSH_TYPE_GLOBAL;
+}
 
+#ifdef CONFIG_SMP
 void radix__flush_tlb_mm(struct mm_struct *mm)
 {
 	unsigned long pid;
+	enum tlb_flush_type type;
 
 	pid = mm->context.id;
 	if (unlikely(pid == MMU_NO_CONTEXT))
@@ -718,16 +757,13 @@ void radix__flush_tlb_mm(struct mm_struct *mm)
 
 	preempt_disable();
 	/*
-	 * Order loads of mm_cpumask vs previous stores to clear ptes before
-	 * the invalidate. See barrier in switch_mm_irqs_off
+	 * Order loads of mm_cpumask (in flush_type_needed) vs previous
+	 * stores to clear ptes before the invalidate. See barrier in
+	 * switch_mm_irqs_off
 	 */
 	smp_mb();
-	if (!mm_is_thread_local(mm)) {
-		if (unlikely(mm_is_singlethreaded(mm))) {
-			exit_flush_lazy_tlbs(mm);
-			goto local;
-		}
-
+	type = flush_type_needed(mm, false);
+	if (type == FLUSH_TYPE_GLOBAL) {
 		if (!mmu_has_feature(MMU_FTR_GTSE)) {
 			unsigned long tgt = H_RPTI_TARGET_CMMU;
 
@@ -744,7 +780,6 @@ void radix__flush_tlb_mm(struct mm_struct *mm)
 			_tlbiel_pid_multicast(mm, pid, RIC_FLUSH_TLB);
 		}
 	} else {
-local:
 		_tlbiel_pid(pid, RIC_FLUSH_TLB);
 	}
 	preempt_enable();
@@ -754,6 +789,7 @@ EXPORT_SYMBOL(radix__flush_tlb_mm);
 static void __flush_all_mm(struct mm_struct *mm, bool fullmm)
 {
 	unsigned long pid;
+	enum tlb_flush_type type;
 
 	pid = mm->context.id;
 	if (unlikely(pid == MMU_NO_CONTEXT))
@@ -761,13 +797,8 @@ static void __flush_all_mm(struct mm_struct *mm, bool fullmm)
 
 	preempt_disable();
 	smp_mb(); /* see radix__flush_tlb_mm */
-	if (!mm_is_thread_local(mm)) {
-		if (unlikely(mm_is_singlethreaded(mm))) {
-			if (!fullmm) {
-				exit_flush_lazy_tlbs(mm);
-				goto local;
-			}
-		}
+	type = flush_type_needed(mm, fullmm);
+	if (type == FLUSH_TYPE_GLOBAL) {
 		if (!mmu_has_feature(MMU_FTR_GTSE)) {
 			unsigned long tgt = H_RPTI_TARGET_CMMU;
 			unsigned long type = H_RPTI_TYPE_TLB | H_RPTI_TYPE_PWC |
@@ -782,7 +813,6 @@ static void __flush_all_mm(struct mm_struct *mm, bool fullmm)
 		else
 			_tlbiel_pid_multicast(mm, pid, RIC_FLUSH_ALL);
 	} else {
-local:
 		_tlbiel_pid(pid, RIC_FLUSH_ALL);
 	}
 	preempt_enable();
@@ -798,6 +828,7 @@ void radix__flush_tlb_page_psize(struct mm_struct *mm, unsigned long vmaddr,
 				 int psize)
 {
 	unsigned long pid;
+	enum tlb_flush_type type;
 
 	pid = mm->context.id;
 	if (unlikely(pid == MMU_NO_CONTEXT))
@@ -805,11 +836,8 @@ void radix__flush_tlb_page_psize(struct mm_struct *mm, unsigned long vmaddr,
 
 	preempt_disable();
 	smp_mb(); /* see radix__flush_tlb_mm */
-	if (!mm_is_thread_local(mm)) {
-		if (unlikely(mm_is_singlethreaded(mm))) {
-			exit_flush_lazy_tlbs(mm);
-			goto local;
-		}
+	type = flush_type_needed(mm, false);
+	if (type == FLUSH_TYPE_GLOBAL) {
 		if (!mmu_has_feature(MMU_FTR_GTSE)) {
 			unsigned long tgt, pg_sizes, size;
 
@@ -827,7 +855,6 @@ void radix__flush_tlb_page_psize(struct mm_struct *mm, unsigned long vmaddr,
 		else
 			_tlbiel_va_multicast(mm, vmaddr, pid, psize, RIC_FLUSH_TLB);
 	} else {
-local:
 		_tlbiel_va(vmaddr, pid, psize, RIC_FLUSH_TLB);
 	}
 	preempt_enable();
@@ -843,8 +870,6 @@ void radix__flush_tlb_page(struct vm_area_struct *vma, unsigned long vmaddr)
 }
 EXPORT_SYMBOL(radix__flush_tlb_page);
 
-#else /* CONFIG_SMP */
-static inline void exit_flush_lazy_tlbs(struct mm_struct *mm) { }
 #endif /* CONFIG_SMP */
 
 static void do_tlbiel_kernel(void *info)
@@ -908,7 +933,9 @@ static inline void __radix__flush_tlb_range(struct mm_struct *mm,
 	unsigned int page_shift = mmu_psize_defs[mmu_virtual_psize].shift;
 	unsigned long page_size = 1UL << page_shift;
 	unsigned long nr_pages = (end - start) >> page_shift;
-	bool local, full;
+	bool fullmm = (end == TLB_FLUSH_ALL);
+	bool flush_pid;
+	enum tlb_flush_type type;
 
 	pid = mm->context.id;
 	if (unlikely(pid == MMU_NO_CONTEXT))
@@ -916,24 +943,16 @@ static inline void __radix__flush_tlb_range(struct mm_struct *mm,
 
 	preempt_disable();
 	smp_mb(); /* see radix__flush_tlb_mm */
-	if (!mm_is_thread_local(mm)) {
-		if (unlikely(mm_is_singlethreaded(mm))) {
-			if (end != TLB_FLUSH_ALL) {
-				exit_flush_lazy_tlbs(mm);
-				goto is_local;
-			}
-		}
-		local = false;
-		full = (end == TLB_FLUSH_ALL ||
-				nr_pages > tlb_single_page_flush_ceiling);
-	} else {
-is_local:
-		local = true;
-		full = (end == TLB_FLUSH_ALL ||
-				nr_pages > tlb_local_single_page_flush_ceiling);
-	}
+	type = flush_type_needed(mm, fullmm);
+
+	if (fullmm)
+		flush_pid = true;
+	else if (type == FLUSH_TYPE_GLOBAL)
+		flush_pid = nr_pages > tlb_single_page_flush_ceiling;
+	else
+		flush_pid = nr_pages > tlb_local_single_page_flush_ceiling;
 
-	if (!mmu_has_feature(MMU_FTR_GTSE) && !local) {
+	if (!mmu_has_feature(MMU_FTR_GTSE) && type == FLUSH_TYPE_GLOBAL) {
 		unsigned long tgt = H_RPTI_TARGET_CMMU;
 		unsigned long pg_sizes = psize_to_rpti_pgsize(mmu_virtual_psize);
 
@@ -943,8 +962,8 @@ static inline void __radix__flush_tlb_range(struct mm_struct *mm,
 			tgt |= H_RPTI_TARGET_NMMU;
 		pseries_rpt_invalidate(pid, tgt, H_RPTI_TYPE_TLB, pg_sizes,
 				       start, end);
-	} else if (full) {
-		if (local) {
+	} else if (flush_pid) {
+		if (type == FLUSH_TYPE_LOCAL) {
 			_tlbiel_pid(pid, RIC_FLUSH_TLB);
 		} else {
 			if (cputlb_use_tlbie()) {
@@ -967,7 +986,7 @@ static inline void __radix__flush_tlb_range(struct mm_struct *mm,
 				hflush = true;
 		}
 
-		if (local) {
+		if (type == FLUSH_TYPE_LOCAL) {
 			asm volatile("ptesync": : :"memory");
 			__tlbiel_va_range(start, end, pid, page_size, mmu_virtual_psize);
 			if (hflush)
@@ -1100,32 +1119,28 @@ static __always_inline void __radix__flush_tlb_range_psize(struct mm_struct *mm,
 	unsigned int page_shift = mmu_psize_defs[psize].shift;
 	unsigned long page_size = 1UL << page_shift;
 	unsigned long nr_pages = (end - start) >> page_shift;
-	bool local, full;
+	bool fullmm = (end == TLB_FLUSH_ALL);
+	bool flush_pid;
+	enum tlb_flush_type type;
 
 	pid = mm->context.id;
 	if (unlikely(pid == MMU_NO_CONTEXT))
 		return;
 
+	fullmm = (end == TLB_FLUSH_ALL);
+
 	preempt_disable();
 	smp_mb(); /* see radix__flush_tlb_mm */
-	if (!mm_is_thread_local(mm)) {
-		if (unlikely(mm_is_singlethreaded(mm))) {
-			if (end != TLB_FLUSH_ALL) {
-				exit_flush_lazy_tlbs(mm);
-				goto is_local;
-			}
-		}
-		local = false;
-		full = (end == TLB_FLUSH_ALL ||
-				nr_pages > tlb_single_page_flush_ceiling);
-	} else {
-is_local:
-		local = true;
-		full = (end == TLB_FLUSH_ALL ||
-				nr_pages > tlb_local_single_page_flush_ceiling);
-	}
+	type = flush_type_needed(mm, fullmm);
 
-	if (!mmu_has_feature(MMU_FTR_GTSE) && !local) {
+	if (fullmm)
+		flush_pid = true;
+	else if (type == FLUSH_TYPE_GLOBAL)
+		flush_pid = nr_pages > tlb_single_page_flush_ceiling;
+	else
+		flush_pid = nr_pages > tlb_local_single_page_flush_ceiling;
+
+	if (!mmu_has_feature(MMU_FTR_GTSE) && type == FLUSH_TYPE_GLOBAL) {
 		unsigned long tgt = H_RPTI_TARGET_CMMU;
 		unsigned long type = H_RPTI_TYPE_TLB;
 		unsigned long pg_sizes = psize_to_rpti_pgsize(psize);
@@ -1135,8 +1150,8 @@ static __always_inline void __radix__flush_tlb_range_psize(struct mm_struct *mm,
 		if (atomic_read(&mm->context.copros) > 0)
 			tgt |= H_RPTI_TARGET_NMMU;
 		pseries_rpt_invalidate(pid, tgt, type, pg_sizes, start, end);
-	} else if (full) {
-		if (local) {
+	} else if (flush_pid) {
+		if (type == FLUSH_TYPE_LOCAL) {
 			_tlbiel_pid(pid, also_pwc ? RIC_FLUSH_ALL : RIC_FLUSH_TLB);
 		} else {
 			if (cputlb_use_tlbie()) {
@@ -1152,7 +1167,7 @@ static __always_inline void __radix__flush_tlb_range_psize(struct mm_struct *mm,
 
 		}
 	} else {
-		if (local)
+		if (type == FLUSH_TYPE_LOCAL)
 			_tlbiel_va_range(start, end, pid, page_size, psize, also_pwc);
 		else if (cputlb_use_tlbie())
 			_tlbie_va_range(start, end, pid, page_size, psize, also_pwc);
@@ -1179,6 +1194,7 @@ static void radix__flush_tlb_pwc_range_psize(struct mm_struct *mm, unsigned long
 void radix__flush_tlb_collapsed_pmd(struct mm_struct *mm, unsigned long addr)
 {
 	unsigned long pid, end;
+	enum tlb_flush_type type;
 
 	pid = mm->context.id;
 	if (unlikely(pid == MMU_NO_CONTEXT))
@@ -1195,11 +1211,8 @@ void radix__flush_tlb_collapsed_pmd(struct mm_struct *mm, unsigned long addr)
 	/* Otherwise first do the PWC, then iterate the pages. */
 	preempt_disable();
 	smp_mb(); /* see radix__flush_tlb_mm */
-	if (!mm_is_thread_local(mm)) {
-		if (unlikely(mm_is_singlethreaded(mm))) {
-			exit_flush_lazy_tlbs(mm);
-			goto local;
-		}
+	type = flush_type_needed(mm, false);
+	if (type == FLUSH_TYPE_GLOBAL) {
 		if (!mmu_has_feature(MMU_FTR_GTSE)) {
 			unsigned long tgt, type, pg_sizes;
 
@@ -1218,7 +1231,6 @@ void radix__flush_tlb_collapsed_pmd(struct mm_struct *mm, unsigned long addr)
 			_tlbiel_va_range_multicast(mm,
 					addr, end, pid, PAGE_SIZE, mmu_virtual_psize, true);
 	} else {
-local:
 		_tlbiel_va_range(addr, end, pid, PAGE_SIZE, mmu_virtual_psize, true);
 	}