Merge branch 'fixes' of git://git.linaro.org/people/rmk/linux-arm

Pull ARM fixes from Russell King:
 "Another round of ARM fixes, which include:
   - Fixing a problem with LPAE mapping sections
   - Reporting of some hwcaps on Krait CPUs
   - Avoiding repetitive warnings in the breakpoint code
   - Fixing a build error noticed on Dove platforms with PJ4 CPUs
   - Fix masking of level 2 cache revision.
   - Fixing timer-based udelay()
   - A larger fix for an erratum causing people major grief with Cortex
     A15 CPUs"

* 'fixes' of git://git.linaro.org/people/rmk/linux-arm:
  ARM: 7690/1: mm: fix CONFIG_LPAE typos
  ARM: 7689/1: add unwind annotations to ftrace asm
  ARM: 7685/1: delay: use private ticks_per_jiffy field for timer-based delay ops
  ARM: 7684/1: errata: Workaround for Cortex-A15 erratum 798181 (TLBI/DSB operations)
  ARM: 7682/1: cache-l2x0: fix masking of RTL revision numbering and set_debug init
  ARM: iWMMXt: always enable iWMMXt support with PJ4 CPUs
  ARM: 7681/1: hw_breakpoint: use warn_once to avoid spam from reset_ctrl_regs()
  ARM: 7678/1: Work around faulty ISAR0 register in some Krait CPUs
  ARM: 7680/1: Detect support for SDIV/UDIV from ISAR0 register
  ARM: 7679/1: Clear IDIVT hwcap if CONFIG_ARM_THUMB=n
  ARM: 7677/1: LPAE: Fix mapping in alloc_init_section for unaligned addresses
  ARM: KVM: vgic: take distributor lock on sync_hwstate path
  ARM: KVM: vgic: force EOIed LRs to the empty state

arch/arm/Kconfig

@@ -1183,9 +1183,9 @@ config ARM_NR_BANKS
 	default 8
 
 config IWMMXT
-	bool "Enable iWMMXt support"
+	bool "Enable iWMMXt support" if !CPU_PJ4
 	depends on CPU_XSCALE || CPU_XSC3 || CPU_MOHAWK || CPU_PJ4
-	default y if PXA27x || PXA3xx || ARCH_MMP
+	default y if PXA27x || PXA3xx || ARCH_MMP || CPU_PJ4
 	help
 	  Enable support for iWMMXt context switching at run time if
 	  running on a CPU that supports it.
@@ -1439,6 +1439,16 @@ config ARM_ERRATA_775420
 	 to deadlock. This workaround puts DSB before executing ISB if
 	 an abort may occur on cache maintenance.
 
+config ARM_ERRATA_798181
+	bool "ARM errata: TLBI/DSB failure on Cortex-A15"
+	depends on CPU_V7 && SMP
+	help
+	  On Cortex-A15 (r0p0..r3p2) the TLBI*IS/DSB operations are not
+	  adequately shooting down all use of the old entries. This
+	  option enables the Linux kernel workaround for this erratum
+	  which sends an IPI to the CPUs that are running the same ASID
+	  as the one being invalidated.
+
 endmenu
 
 source "arch/arm/common/Kconfig"

arch/arm/include/asm/delay.h

@@ -24,7 +24,7 @@ extern struct arm_delay_ops {
 	void (*delay)(unsigned long);
 	void (*const_udelay)(unsigned long);
 	void (*udelay)(unsigned long);
-	bool const_clock;
+	unsigned long ticks_per_jiffy;
 } arm_delay_ops;
 
 #define __delay(n)		arm_delay_ops.delay(n)

arch/arm/include/asm/highmem.h

@@ -41,6 +41,13 @@ extern void kunmap_high(struct page *page);
 #endif
 #endif
 
+/*
+ * Needed to be able to broadcast the TLB invalidation for kmap.
+ */
+#ifdef CONFIG_ARM_ERRATA_798181
+#undef ARCH_NEEDS_KMAP_HIGH_GET
+#endif
+
 #ifdef ARCH_NEEDS_KMAP_HIGH_GET
 extern void *kmap_high_get(struct page *page);
 #else

arch/arm/include/asm/mmu_context.h

@@ -27,6 +27,8 @@ void __check_vmalloc_seq(struct mm_struct *mm);
 void check_and_switch_context(struct mm_struct *mm, struct task_struct *tsk);
 #define init_new_context(tsk,mm)	({ atomic64_set(&mm->context.id, 0); 0; })
 
+DECLARE_PER_CPU(atomic64_t, active_asids);
+
 #else	/* !CONFIG_CPU_HAS_ASID */
 
 #ifdef CONFIG_MMU

arch/arm/include/asm/tlbflush.h

@@ -450,6 +450,21 @@ static inline void local_flush_bp_all(void)
 		isb();
 }
 
+#ifdef CONFIG_ARM_ERRATA_798181
+static inline void dummy_flush_tlb_a15_erratum(void)
+{
+	/*
+	 * Dummy TLBIMVAIS. Using the unmapped address 0 and ASID 0.
+	 */
+	asm("mcr p15, 0, %0, c8, c3, 1" : : "r" (0));
+	dsb();
+}
+#else
+static inline void dummy_flush_tlb_a15_erratum(void)
+{
+}
+#endif
+
 /*
  *	flush_pmd_entry
  *

arch/arm/kernel/entry-common.S

@@ -276,7 +276,13 @@ ENDPROC(ftrace_graph_caller_old)
  */
 
 .macro mcount_enter
+/*
+ * This pad compensates for the push {lr} at the call site.  Note that we are
+ * unable to unwind through a function which does not otherwise save its lr.
+ */
+ UNWIND(.pad	#4)
 	stmdb	sp!, {r0-r3, lr}
+ UNWIND(.save	{r0-r3, lr})
 .endm
 
 .macro mcount_get_lr reg
@@ -289,6 +295,7 @@ ENDPROC(ftrace_graph_caller_old)
 .endm
 
 ENTRY(__gnu_mcount_nc)
+UNWIND(.fnstart)
 #ifdef CONFIG_DYNAMIC_FTRACE
 	mov	ip, lr
 	ldmia	sp!, {lr}
@@ -296,17 +303,22 @@ ENTRY(__gnu_mcount_nc)
 #else
 	__mcount
 #endif
+UNWIND(.fnend)
 ENDPROC(__gnu_mcount_nc)
 
 #ifdef CONFIG_DYNAMIC_FTRACE
 ENTRY(ftrace_caller)
+UNWIND(.fnstart)
 	__ftrace_caller
+UNWIND(.fnend)
 ENDPROC(ftrace_caller)
 #endif
 
 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
 ENTRY(ftrace_graph_caller)
+UNWIND(.fnstart)
 	__ftrace_graph_caller
+UNWIND(.fnend)
 ENDPROC(ftrace_graph_caller)
 #endif
 

arch/arm/kernel/head.S

@@ -267,7 +267,7 @@ __create_page_tables:
 	addne	r6, r6, #1 << SECTION_SHIFT
 	strne	r6, [r3]
 
-#if defined(CONFIG_LPAE) && defined(CONFIG_CPU_ENDIAN_BE8)
+#if defined(CONFIG_ARM_LPAE) && defined(CONFIG_CPU_ENDIAN_BE8)
 	sub	r4, r4, #4			@ Fixup page table pointer
 						@ for 64-bit descriptors
 #endif

arch/arm/kernel/hw_breakpoint.c

@@ -966,7 +966,7 @@ static void reset_ctrl_regs(void *unused)
 	}
 
 	if (err) {
-		pr_warning("CPU %d debug is powered down!\n", cpu);
+		pr_warn_once("CPU %d debug is powered down!\n", cpu);
 		cpumask_or(&debug_err_mask, &debug_err_mask, cpumask_of(cpu));
 		return;
 	}
@@ -987,7 +987,7 @@ static void reset_ctrl_regs(void *unused)
 	isb();
 
 	if (cpumask_intersects(&debug_err_mask, cpumask_of(cpu))) {
-		pr_warning("CPU %d failed to disable vector catch\n", cpu);
+		pr_warn_once("CPU %d failed to disable vector catch\n", cpu);
 		return;
 	}
 
@@ -1007,7 +1007,7 @@ static void reset_ctrl_regs(void *unused)
 	}
 
 	if (cpumask_intersects(&debug_err_mask, cpumask_of(cpu))) {
-		pr_warning("CPU %d failed to clear debug register pairs\n", cpu);
+		pr_warn_once("CPU %d failed to clear debug register pairs\n", cpu);
 		return;
 	}
 

arch/arm/kernel/setup.c

@@ -353,6 +353,23 @@ void __init early_print(const char *str, ...)
 	printk("%s", buf);
 }
 
+static void __init cpuid_init_hwcaps(void)
+{
+	unsigned int divide_instrs;
+
+	if (cpu_architecture() < CPU_ARCH_ARMv7)
+		return;
+
+	divide_instrs = (read_cpuid_ext(CPUID_EXT_ISAR0) & 0x0f000000) >> 24;
+
+	switch (divide_instrs) {
+	case 2:
+		elf_hwcap |= HWCAP_IDIVA;
+	case 1:
+		elf_hwcap |= HWCAP_IDIVT;
+	}
+}
+
 static void __init feat_v6_fixup(void)
 {
 	int id = read_cpuid_id();
@@ -483,8 +500,11 @@ static void __init setup_processor(void)
 	snprintf(elf_platform, ELF_PLATFORM_SIZE, "%s%c",
 		 list->elf_name, ENDIANNESS);
 	elf_hwcap = list->elf_hwcap;
+
+	cpuid_init_hwcaps();
+
 #ifndef CONFIG_ARM_THUMB
-	elf_hwcap &= ~HWCAP_THUMB;
+	elf_hwcap &= ~(HWCAP_THUMB | HWCAP_IDIVT);
 #endif
 
 	feat_v6_fixup();
@@ -524,7 +544,7 @@ int __init arm_add_memory(phys_addr_t start, phys_addr_t size)
 	size -= start & ~PAGE_MASK;
 	bank->start = PAGE_ALIGN(start);
 
-#ifndef CONFIG_LPAE
+#ifndef CONFIG_ARM_LPAE
 	if (bank->start + size < bank->start) {
 		printk(KERN_CRIT "Truncating memory at 0x%08llx to fit in "
 			"32-bit physical address space\n", (long long)start);

arch/arm/kernel/smp.c

@@ -673,9 +673,6 @@ static int cpufreq_callback(struct notifier_block *nb,
 	if (freq->flags & CPUFREQ_CONST_LOOPS)
 		return NOTIFY_OK;
 
-	if (arm_delay_ops.const_clock)
-		return NOTIFY_OK;
-
 	if (!per_cpu(l_p_j_ref, cpu)) {
 		per_cpu(l_p_j_ref, cpu) =
 			per_cpu(cpu_data, cpu).loops_per_jiffy;

arch/arm/kernel/smp_tlb.c

@@ -12,6 +12,7 @@
 
 #include <asm/smp_plat.h>
 #include <asm/tlbflush.h>
+#include <asm/mmu_context.h>
 
 /**********************************************************************/
 
@@ -69,12 +70,72 @@ static inline void ipi_flush_bp_all(void *ignored)
 	local_flush_bp_all();
 }
 
+#ifdef CONFIG_ARM_ERRATA_798181
+static int erratum_a15_798181(void)
+{
+	unsigned int midr = read_cpuid_id();
+
+	/* Cortex-A15 r0p0..r3p2 affected */
+	if ((midr & 0xff0ffff0) != 0x410fc0f0 || midr > 0x413fc0f2)
+		return 0;
+	return 1;
+}
+#else
+static int erratum_a15_798181(void)
+{
+	return 0;
+}
+#endif
+
+static void ipi_flush_tlb_a15_erratum(void *arg)
+{
+	dmb();
+}
+
+static void broadcast_tlb_a15_erratum(void)
+{
+	if (!erratum_a15_798181())
+		return;
+
+	dummy_flush_tlb_a15_erratum();
+	smp_call_function_many(cpu_online_mask, ipi_flush_tlb_a15_erratum,
+			       NULL, 1);
+}
+
+static void broadcast_tlb_mm_a15_erratum(struct mm_struct *mm)
+{
+	int cpu;
+	cpumask_t mask = { CPU_BITS_NONE };
+
+	if (!erratum_a15_798181())
+		return;
+
+	dummy_flush_tlb_a15_erratum();
+	for_each_online_cpu(cpu) {
+		if (cpu == smp_processor_id())
+			continue;
+		/*
+		 * We only need to send an IPI if the other CPUs are running
+		 * the same ASID as the one being invalidated. There is no
+		 * need for locking around the active_asids check since the
+		 * switch_mm() function has at least one dmb() (as required by
+		 * this workaround) in case a context switch happens on
+		 * another CPU after the condition below.
+		 */
+		if (atomic64_read(&mm->context.id) ==
+		    atomic64_read(&per_cpu(active_asids, cpu)))
+			cpumask_set_cpu(cpu, &mask);
+	}
+	smp_call_function_many(&mask, ipi_flush_tlb_a15_erratum, NULL, 1);
+}
+
 void flush_tlb_all(void)
 {
 	if (tlb_ops_need_broadcast())
 		on_each_cpu(ipi_flush_tlb_all, NULL, 1);
 	else
 		local_flush_tlb_all();
+	broadcast_tlb_a15_erratum();
 }
 
 void flush_tlb_mm(struct mm_struct *mm)
@@ -83,6 +144,7 @@ void flush_tlb_mm(struct mm_struct *mm)
 		on_each_cpu_mask(mm_cpumask(mm), ipi_flush_tlb_mm, mm, 1);
 	else
 		local_flush_tlb_mm(mm);
+	broadcast_tlb_mm_a15_erratum(mm);
 }
 
 void flush_tlb_page(struct vm_area_struct *vma, unsigned long uaddr)
@@ -95,6 +157,7 @@ void flush_tlb_page(struct vm_area_struct *vma, unsigned long uaddr)
 					&ta, 1);
 	} else
 		local_flush_tlb_page(vma, uaddr);
+	broadcast_tlb_mm_a15_erratum(vma->vm_mm);
 }
 
 void flush_tlb_kernel_page(unsigned long kaddr)
@@ -105,6 +168,7 @@ void flush_tlb_kernel_page(unsigned long kaddr)
 		on_each_cpu(ipi_flush_tlb_kernel_page, &ta, 1);
 	} else
 		local_flush_tlb_kernel_page(kaddr);
+	broadcast_tlb_a15_erratum();
 }
 
 void flush_tlb_range(struct vm_area_struct *vma,
@@ -119,6 +183,7 @@ void flush_tlb_range(struct vm_area_struct *vma,
 					&ta, 1);
 	} else
 		local_flush_tlb_range(vma, start, end);
+	broadcast_tlb_mm_a15_erratum(vma->vm_mm);
 }
 
 void flush_tlb_kernel_range(unsigned long start, unsigned long end)
@@ -130,6 +195,7 @@ void flush_tlb_kernel_range(unsigned long start, unsigned long end)
 		on_each_cpu(ipi_flush_tlb_kernel_range, &ta, 1);
 	} else
 		local_flush_tlb_kernel_range(start, end);
+	broadcast_tlb_a15_erratum();
 }
 
 void flush_bp_all(void)

arch/arm/kvm/vgic.c

@@ -883,8 +883,7 @@ static bool vgic_queue_irq(struct kvm_vcpu *vcpu, u8 sgi_source_id, int irq)
 			  lr, irq, vgic_cpu->vgic_lr[lr]);
 		BUG_ON(!test_bit(lr, vgic_cpu->lr_used));
 		vgic_cpu->vgic_lr[lr] |= GICH_LR_PENDING_BIT;
-
-		goto out;
+		return true;
 	}
 
 	/* Try to use another LR for this interrupt */
@@ -898,7 +897,6 @@ static bool vgic_queue_irq(struct kvm_vcpu *vcpu, u8 sgi_source_id, int irq)
 	vgic_cpu->vgic_irq_lr_map[irq] = lr;
 	set_bit(lr, vgic_cpu->lr_used);
 
-out:
 	if (!vgic_irq_is_edge(vcpu, irq))
 		vgic_cpu->vgic_lr[lr] |= GICH_LR_EOI;
 
@@ -1018,21 +1016,6 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
 
 	kvm_debug("MISR = %08x\n", vgic_cpu->vgic_misr);
 
-	/*
-	 * We do not need to take the distributor lock here, since the only
-	 * action we perform is clearing the irq_active_bit for an EOIed
-	 * level interrupt.  There is a potential race with
-	 * the queuing of an interrupt in __kvm_vgic_flush_hwstate(), where we
-	 * check if the interrupt is already active. Two possibilities:
-	 *
-	 * - The queuing is occurring on the same vcpu: cannot happen,
-	 *   as we're already in the context of this vcpu, and
-	 *   executing the handler
-	 * - The interrupt has been migrated to another vcpu, and we
-	 *   ignore this interrupt for this run. Big deal. It is still
-	 *   pending though, and will get considered when this vcpu
-	 *   exits.
-	 */
 	if (vgic_cpu->vgic_misr & GICH_MISR_EOI) {
 		/*
 		 * Some level interrupts have been EOIed. Clear their
@@ -1054,6 +1037,13 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
 			} else {
 				vgic_cpu_irq_clear(vcpu, irq);
 			}
+
+			/*
+			 * Despite being EOIed, the LR may not have
+			 * been marked as empty.
+			 */
+			set_bit(lr, (unsigned long *)vgic_cpu->vgic_elrsr);
+			vgic_cpu->vgic_lr[lr] &= ~GICH_LR_ACTIVE_BIT;
 		}
 	}
 
@@ -1064,9 +1054,8 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
 }
 
 /*
- * Sync back the VGIC state after a guest run. We do not really touch
- * the distributor here (the irq_pending_on_cpu bit is safe to set),
- * so there is no need for taking its lock.
+ * Sync back the VGIC state after a guest run. The distributor lock is
+ * needed so we don't get preempted in the middle of the state processing.
  */
 static void __kvm_vgic_sync_hwstate(struct kvm_vcpu *vcpu)
 {
@@ -1112,10 +1101,14 @@ void kvm_vgic_flush_hwstate(struct kvm_vcpu *vcpu)
 
 void kvm_vgic_sync_hwstate(struct kvm_vcpu *vcpu)
 {
+	struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
+
 	if (!irqchip_in_kernel(vcpu->kvm))
 		return;
 
+	spin_lock(&dist->lock);
 	__kvm_vgic_sync_hwstate(vcpu);
+	spin_unlock(&dist->lock);
 }
 
 int kvm_vgic_vcpu_pending_irq(struct kvm_vcpu *vcpu)

arch/arm/lib/delay.c

@@ -58,7 +58,7 @@ static void __timer_delay(unsigned long cycles)
 static void __timer_const_udelay(unsigned long xloops)
 {
 	unsigned long long loops = xloops;
-	loops *= loops_per_jiffy;
+	loops *= arm_delay_ops.ticks_per_jiffy;
 	__timer_delay(loops >> UDELAY_SHIFT);
 }
 
@@ -73,11 +73,13 @@ void __init register_current_timer_delay(const struct delay_timer *timer)
 		pr_info("Switching to timer-based delay loop\n");
 		delay_timer			= timer;
 		lpj_fine			= timer->freq / HZ;
-		loops_per_jiffy			= lpj_fine;
+
+		/* cpufreq may scale loops_per_jiffy, so keep a private copy */
+		arm_delay_ops.ticks_per_jiffy	= lpj_fine;
 		arm_delay_ops.delay		= __timer_delay;
 		arm_delay_ops.const_udelay	= __timer_const_udelay;
 		arm_delay_ops.udelay		= __timer_udelay;
-		arm_delay_ops.const_clock	= true;
+
 		delay_calibrated		= true;
 	} else {
 		pr_info("Ignoring duplicate/late registration of read_current_timer delay\n");

arch/arm/mm/cache-l2x0.c

@@ -299,7 +299,7 @@ static void l2x0_unlock(u32 cache_id)
 	int lockregs;
 	int i;
 
-	switch (cache_id) {
+	switch (cache_id & L2X0_CACHE_ID_PART_MASK) {
 	case L2X0_CACHE_ID_PART_L310:
 		lockregs = 8;
 		break;
@@ -333,15 +333,14 @@ void __init l2x0_init(void __iomem *base, u32 aux_val, u32 aux_mask)
 	if (cache_id_part_number_from_dt)
 		cache_id = cache_id_part_number_from_dt;
 	else
-		cache_id = readl_relaxed(l2x0_base + L2X0_CACHE_ID)
-			& L2X0_CACHE_ID_PART_MASK;
+		cache_id = readl_relaxed(l2x0_base + L2X0_CACHE_ID);
 	aux = readl_relaxed(l2x0_base + L2X0_AUX_CTRL);
 
 	aux &= aux_mask;
 	aux |= aux_val;
 
 	/* Determine the number of ways */
-	switch (cache_id) {
+	switch (cache_id & L2X0_CACHE_ID_PART_MASK) {
 	case L2X0_CACHE_ID_PART_L310:
 		if (aux & (1 << 16))
 			ways = 16;
@@ -725,7 +724,6 @@ static const struct l2x0_of_data pl310_data = {
 		.flush_all   = l2x0_flush_all,
 		.inv_all     = l2x0_inv_all,
 		.disable     = l2x0_disable,
-		.set_debug   = pl310_set_debug,
 	},
 };
 
@@ -814,9 +812,8 @@ int __init l2x0_of_init(u32 aux_val, u32 aux_mask)
 		data->save();
 
 	of_init = true;
-	l2x0_init(l2x0_base, aux_val, aux_mask);
-
 	memcpy(&outer_cache, &data->outer_cache, sizeof(outer_cache));
+	l2x0_init(l2x0_base, aux_val, aux_mask);
 
 	return 0;
 }

arch/arm/mm/context.c

@@ -48,7 +48,7 @@ static DEFINE_RAW_SPINLOCK(cpu_asid_lock);
 static atomic64_t asid_generation = ATOMIC64_INIT(ASID_FIRST_VERSION);
 static DECLARE_BITMAP(asid_map, NUM_USER_ASIDS);
 
-static DEFINE_PER_CPU(atomic64_t, active_asids);
+DEFINE_PER_CPU(atomic64_t, active_asids);
 static DEFINE_PER_CPU(u64, reserved_asids);
 static cpumask_t tlb_flush_pending;
 
@@ -215,6 +215,7 @@ void check_and_switch_context(struct mm_struct *mm, struct task_struct *tsk)
 	if (cpumask_test_and_clear_cpu(cpu, &tlb_flush_pending)) {
 		local_flush_bp_all();
 		local_flush_tlb_all();
+		dummy_flush_tlb_a15_erratum();
 	}
 
 	atomic64_set(&per_cpu(active_asids, cpu), asid);

arch/arm/mm/mmu.c

@@ -598,39 +598,60 @@ static void __init alloc_init_pte(pmd_t *pmd, unsigned long addr,
 	} while (pte++, addr += PAGE_SIZE, addr != end);
 }
 
-static void __init alloc_init_section(pud_t *pud, unsigned long addr,
-				      unsigned long end, phys_addr_t phys,
-				      const struct mem_type *type)
+static void __init map_init_section(pmd_t *pmd, unsigned long addr,
+			unsigned long end, phys_addr_t phys,
+			const struct mem_type *type)
 {
-	pmd_t *pmd = pmd_offset(pud, addr);
-
+#ifndef CONFIG_ARM_LPAE
 	/*
-	 * Try a section mapping - end, addr and phys must all be aligned
-	 * to a section boundary.  Note that PMDs refer to the individual
-	 * L1 entries, whereas PGDs refer to a group of L1 entries making
-	 * up one logical pointer to an L2 table.
+	 * In classic MMU format, puds and pmds are folded in to
+	 * the pgds. pmd_offset gives the PGD entry. PGDs refer to a
+	 * group of L1 entries making up one logical pointer to
+	 * an L2 table (2MB), where as PMDs refer to the individual
+	 * L1 entries (1MB). Hence increment to get the correct
+	 * offset for odd 1MB sections.
+	 * (See arch/arm/include/asm/pgtable-2level.h)
 	 */
-	if (type->prot_sect && ((addr | end | phys) & ~SECTION_MASK) == 0) {
-		pmd_t *p = pmd;
-
-#ifndef CONFIG_ARM_LPAE
-		if (addr & SECTION_SIZE)
-			pmd++;
+	if (addr & SECTION_SIZE)
+		pmd++;
 #endif
+	do {
+		*pmd = __pmd(phys | type->prot_sect);
+		phys += SECTION_SIZE;
+	} while (pmd++, addr += SECTION_SIZE, addr != end);
 
-		do {
-			*pmd = __pmd(phys | type->prot_sect);
-			phys += SECTION_SIZE;
-		} while (pmd++, addr += SECTION_SIZE, addr != end);
+	flush_pmd_entry(pmd);
+}
 
-		flush_pmd_entry(p);
-	} else {
+static void __init alloc_init_pmd(pud_t *pud, unsigned long addr,
+				      unsigned long end, phys_addr_t phys,
+				      const struct mem_type *type)
+{
+	pmd_t *pmd = pmd_offset(pud, addr);
+	unsigned long next;
+
+	do {
 		/*
-		 * No need to loop; pte's aren't interested in the
-		 * individual L1 entries.
+		 * With LPAE, we must loop over to map
+		 * all the pmds for the given range.
 		 */
-		alloc_init_pte(pmd, addr, end, __phys_to_pfn(phys), type);
-	}
+		next = pmd_addr_end(addr, end);
+
+		/*
+		 * Try a section mapping - addr, next and phys must all be
+		 * aligned to a section boundary.
+		 */
+		if (type->prot_sect &&
+				((addr | next | phys) & ~SECTION_MASK) == 0) {
+			map_init_section(pmd, addr, next, phys, type);
+		} else {
+			alloc_init_pte(pmd, addr, next,
+						__phys_to_pfn(phys), type);
+		}
+
+		phys += next - addr;
+
+	} while (pmd++, addr = next, addr != end);
 }
 
 static void __init alloc_init_pud(pgd_t *pgd, unsigned long addr,
@@ -641,7 +662,7 @@ static void __init alloc_init_pud(pgd_t *pgd, unsigned long addr,
 
 	do {
 		next = pud_addr_end(addr, end);
-		alloc_init_section(pud, addr, next, phys, type);
+		alloc_init_pmd(pud, addr, next, phys, type);
 		phys += next - addr;
 	} while (pud++, addr = next, addr != end);
 }

arch/arm/mm/proc-v7.S

@@ -420,7 +420,7 @@ __v7_pj4b_proc_info:
 __v7_ca7mp_proc_info:
 	.long	0x410fc070
 	.long	0xff0ffff0
-	__v7_proc __v7_ca7mp_setup, hwcaps = HWCAP_IDIV
+	__v7_proc __v7_ca7mp_setup
 	.size	__v7_ca7mp_proc_info, . - __v7_ca7mp_proc_info
 
 	/*
@@ -430,9 +430,24 @@ __v7_ca7mp_proc_info:
 __v7_ca15mp_proc_info:
 	.long	0x410fc0f0
 	.long	0xff0ffff0
-	__v7_proc __v7_ca15mp_setup, hwcaps = HWCAP_IDIV
+	__v7_proc __v7_ca15mp_setup
 	.size	__v7_ca15mp_proc_info, . - __v7_ca15mp_proc_info
 
+	/*
+	 * Qualcomm Inc. Krait processors.
+	 */
+	.type	__krait_proc_info, #object
+__krait_proc_info:
+	.long	0x510f0400		@ Required ID value
+	.long	0xff0ffc00		@ Mask for ID
+	/*
+	 * Some Krait processors don't indicate support for SDIV and UDIV
+	 * instructions in the ARM instruction set, even though they actually
+	 * do support them.
+	 */
+	__v7_proc __v7_setup, hwcaps = HWCAP_IDIV
+	.size	__krait_proc_info, . - __krait_proc_info
+
 	/*
 	 * Match any ARMv7 processor core.
 	 */