Merge tag 'for-3.5' of git://openrisc.net/jonas/linux

Pull OpenRISC updates from Jonas Bonn:
 "A couple of cleanups for the OpenRISC architecture:

   - Implement IRQ domains
   - Use DMA mapping framework completely and catch up with recent
     changes to dma_map_ops
   - One bug fix to the "or1k_atomic" syscall to not clobber call-saved
     registers
   - OOM killer patches to the pagefault handler ported from the X86
     arch
   - ...and a couple of header file cleanups"

* tag 'for-3.5' of git://openrisc.net/jonas/linux:
  openrisc: use scratch regs in atomic syscall
  openrisc: provide dma_map_ops
  openrisc: header file cleanups
  openrisc/mm/fault.c: Port OOM changes to do_page_fault
  openrisc: remove unnecessary includes
  openrisc: implement irqdomains

arch/openrisc/Kconfig

@@ -7,6 +7,7 @@ config OPENRISC
 	def_bool y
 	select OF
 	select OF_EARLY_FLATTREE
+	select IRQ_DOMAIN
 	select HAVE_MEMBLOCK
 	select ARCH_WANT_OPTIONAL_GPIOLIB
         select HAVE_ARCH_TRACEHOOK

arch/openrisc/include/asm/Kbuild

 include include/asm-generic/Kbuild.asm
 
-header-y += spr_defs.h
+header-y += elf.h
+header-y += ucontext.h
 
 generic-y += atomic.h
 generic-y += auxvec.h

arch/openrisc/include/asm/dma-mapping.h

@@ -20,150 +20,71 @@
 /*
  * See Documentation/DMA-API-HOWTO.txt and
  * Documentation/DMA-API.txt for documentation.
- *
- * This file is written with the intention of eventually moving over
- * to largely using asm-generic/dma-mapping-common.h in its place.
  */
 
 #include <linux/dma-debug.h>
 #include <asm-generic/dma-coherent.h>
 #include <linux/kmemcheck.h>
+#include <linux/dma-mapping.h>
 
 #define DMA_ERROR_CODE		(~(dma_addr_t)0x0)
 
+extern struct dma_map_ops or1k_dma_map_ops;
 
-#define dma_alloc_noncoherent(d, s, h, f) dma_alloc_coherent(d, s, h, f)
-#define dma_free_noncoherent(d, s, v, h) dma_free_coherent(d, s, v, h)
-
-void *or1k_dma_alloc_coherent(struct device *dev, size_t size,
-			      dma_addr_t *dma_handle, gfp_t flag);
-void or1k_dma_free_coherent(struct device *dev, size_t size, void *vaddr,
-			    dma_addr_t dma_handle);
-dma_addr_t or1k_map_page(struct device *dev, struct page *page,
-			 unsigned long offset, size_t size,
-			 enum dma_data_direction dir,
-			 struct dma_attrs *attrs);
-void or1k_unmap_page(struct device *dev, dma_addr_t dma_handle,
-		     size_t size, enum dma_data_direction dir,
-		     struct dma_attrs *attrs);
-int or1k_map_sg(struct device *dev, struct scatterlist *sg,
-		int nents, enum dma_data_direction dir,
-		struct dma_attrs *attrs);
-void or1k_unmap_sg(struct device *dev, struct scatterlist *sg,
-		   int nents, enum dma_data_direction dir,
-		   struct dma_attrs *attrs);
-void or1k_sync_single_for_cpu(struct device *dev,
-			      dma_addr_t dma_handle, size_t size,
-			      enum dma_data_direction dir);
-void or1k_sync_single_for_device(struct device *dev,
-			         dma_addr_t dma_handle, size_t size,
-			         enum dma_data_direction dir);
-
-static inline void *dma_alloc_coherent(struct device *dev, size_t size,
-					dma_addr_t *dma_handle, gfp_t flag)
+static inline struct dma_map_ops *get_dma_ops(struct device *dev)
 {
-	void *memory;
-
-	memory = or1k_dma_alloc_coherent(dev, size, dma_handle, flag);
-
-	debug_dma_alloc_coherent(dev, size, *dma_handle, memory);
-	return memory;
+	return &or1k_dma_map_ops;
 }
 
-static inline void dma_free_coherent(struct device *dev, size_t size,
-				     void *cpu_addr, dma_addr_t dma_handle)
-{
-	debug_dma_free_coherent(dev, size, cpu_addr, dma_handle);
-	or1k_dma_free_coherent(dev, size, cpu_addr, dma_handle);
-}
+#include <asm-generic/dma-mapping-common.h>
 
-static inline dma_addr_t dma_map_single(struct device *dev, void *ptr,
-					size_t size,
-					enum dma_data_direction dir)
-{
-	dma_addr_t addr;
-
-	kmemcheck_mark_initialized(ptr, size);
-	BUG_ON(!valid_dma_direction(dir));
-	addr = or1k_map_page(dev, virt_to_page(ptr),
-			     (unsigned long)ptr & ~PAGE_MASK, size,
-			     dir, NULL);
-	debug_dma_map_page(dev, virt_to_page(ptr),
-			   (unsigned long)ptr & ~PAGE_MASK, size,
-			   dir, addr, true);
-	return addr;
-}
+#define dma_alloc_coherent(d,s,h,f) dma_alloc_attrs(d,s,h,f,NULL) 
 
-static inline void dma_unmap_single(struct device *dev, dma_addr_t addr,
-					  size_t size,
-					  enum dma_data_direction dir)
+static inline void *dma_alloc_attrs(struct device *dev, size_t size,
+				    dma_addr_t *dma_handle, gfp_t gfp,
+				    struct dma_attrs *attrs)
 {
-	BUG_ON(!valid_dma_direction(dir));
-	or1k_unmap_page(dev, addr, size, dir, NULL);
-	debug_dma_unmap_page(dev, addr, size, dir, true);
-}
+	struct dma_map_ops *ops = get_dma_ops(dev);
+	void *memory;
 
-static inline int dma_map_sg(struct device *dev, struct scatterlist *sg,
-				   int nents, enum dma_data_direction dir)
-{
-	int i, ents;
-	struct scatterlist *s;
+	memory = ops->alloc(dev, size, dma_handle, gfp, attrs);
 
-	for_each_sg(sg, s, nents, i)
-		kmemcheck_mark_initialized(sg_virt(s), s->length);
-	BUG_ON(!valid_dma_direction(dir));
-	ents = or1k_map_sg(dev, sg, nents, dir, NULL);
-	debug_dma_map_sg(dev, sg, nents, ents, dir);
+	debug_dma_alloc_coherent(dev, size, *dma_handle, memory);
 
-	return ents;
+	return memory;
 }
 
-static inline void dma_unmap_sg(struct device *dev, struct scatterlist *sg,
-				      int nents, enum dma_data_direction dir)
-{
-	BUG_ON(!valid_dma_direction(dir));
-	debug_dma_unmap_sg(dev, sg, nents, dir);
-	or1k_unmap_sg(dev, sg, nents, dir, NULL);
-}
+#define dma_free_coherent(d,s,c,h) dma_free_attrs(d,s,c,h,NULL)
 
-static inline dma_addr_t dma_map_page(struct device *dev, struct page *page,
-				      size_t offset, size_t size,
-				      enum dma_data_direction dir)
+static inline void dma_free_attrs(struct device *dev, size_t size,
+				  void *cpu_addr, dma_addr_t dma_handle,
+				  struct dma_attrs *attrs)
 {
-	dma_addr_t addr;
+	struct dma_map_ops *ops = get_dma_ops(dev);
 
-	kmemcheck_mark_initialized(page_address(page) + offset, size);
-	BUG_ON(!valid_dma_direction(dir));
-	addr = or1k_map_page(dev, page, offset, size, dir, NULL);
-	debug_dma_map_page(dev, page, offset, size, dir, addr, false);
+	debug_dma_free_coherent(dev, size, cpu_addr, dma_handle);
 
-	return addr;
+	ops->free(dev, size, cpu_addr, dma_handle, attrs);
 }
 
-static inline void dma_unmap_page(struct device *dev, dma_addr_t addr,
-				  size_t size, enum dma_data_direction dir)
+static inline void *dma_alloc_noncoherent(struct device *dev, size_t size,
+					  dma_addr_t *dma_handle, gfp_t gfp)
 {
-	BUG_ON(!valid_dma_direction(dir));
-	or1k_unmap_page(dev, addr, size, dir, NULL);
-	debug_dma_unmap_page(dev, addr, size, dir, true);
-}
+	struct dma_attrs attrs;
 
-static inline void dma_sync_single_for_cpu(struct device *dev, dma_addr_t addr,
-					   size_t size,
-					   enum dma_data_direction dir)
-{
-	BUG_ON(!valid_dma_direction(dir));
-	or1k_sync_single_for_cpu(dev, addr, size, dir);
-	debug_dma_sync_single_for_cpu(dev, addr, size, dir);
+	dma_set_attr(DMA_ATTR_NON_CONSISTENT, &attrs);
+
+	return dma_alloc_attrs(dev, size, dma_handle, gfp, &attrs);
 }
 
-static inline void dma_sync_single_for_device(struct device *dev,
-					      dma_addr_t addr, size_t size,
-					      enum dma_data_direction dir)
+static inline void dma_free_noncoherent(struct device *dev, size_t size,
+					 void *cpu_addr, dma_addr_t dma_handle)
 {
-	BUG_ON(!valid_dma_direction(dir));
-	or1k_sync_single_for_device(dev, addr, size, dir);
-	debug_dma_sync_single_for_device(dev, addr, size, dir);
+	struct dma_attrs attrs;
+
+	dma_set_attr(DMA_ATTR_NON_CONSISTENT, &attrs);
+
+	dma_free_attrs(dev, size, cpu_addr, dma_handle, &attrs);
 }
 
 static inline int dma_supported(struct device *dev, u64 dma_mask)

arch/openrisc/include/asm/elf.h

@@ -19,12 +19,18 @@
 #ifndef __ASM_OPENRISC_ELF_H
 #define __ASM_OPENRISC_ELF_H
 
+/*
+ * This files is partially exported to userspace.  This allows us to keep
+ * the ELF bits in one place which should assist in keeping the kernel and
+ * userspace in sync.
+ */
+
 /*
  * ELF register definitions..
  */
-#include <linux/types.h>
-#include <linux/ptrace.h>
 
+/* for struct user_regs_struct definition */
+#include <asm/ptrace.h>
 
 /* The OR1K relocation types... not all relevant for module loader */
 #define R_OR32_NONE	0
@@ -62,6 +68,8 @@ typedef unsigned long elf_fpregset_t;
 
 #ifdef __KERNEL__
 
+#include <linux/types.h>
+
 /*
  * This is used to ensure we don't load something for the wrong architecture.
  */

arch/openrisc/include/asm/ptrace.h

@@ -19,8 +19,6 @@
 #ifndef __ASM_OPENRISC_PTRACE_H
 #define __ASM_OPENRISC_PTRACE_H
 
-#include <asm/spr_defs.h>
-
 #ifndef __ASSEMBLY__
 /*
  * This is the layout of the regset returned by the GETREGSET ptrace call
@@ -30,13 +28,13 @@ struct user_regs_struct {
 	unsigned long gpr[32];
 	unsigned long pc;
 	unsigned long sr;
-	unsigned long pad1;
-	unsigned long pad2;
 };
 #endif
 
 #ifdef __KERNEL__
 
+#include <asm/spr_defs.h>
+
 /*
  * Make kernel PTrace/register structures opaque to userspace... userspace can
  * access thread state via the regset mechanism.  This allows us a bit of

arch/openrisc/kernel/dma.c

@@ -21,12 +21,15 @@
 
 #include <linux/dma-mapping.h>
 #include <linux/dma-debug.h>
+#include <linux/export.h>
+#include <linux/dma-attrs.h>
 
 #include <asm/cpuinfo.h>
 #include <asm/spr_defs.h>
 #include <asm/tlbflush.h>
 
-static int page_set_nocache(pte_t *pte, unsigned long addr,
+static int
+page_set_nocache(pte_t *pte, unsigned long addr,
 		 unsigned long next, struct mm_walk *walk)
 {
 	unsigned long cl;
@@ -46,7 +49,8 @@ static int page_set_nocache(pte_t *pte, unsigned long addr,
 	return 0;
 }
 
-static int page_clear_nocache(pte_t *pte, unsigned long addr,
+static int
+page_clear_nocache(pte_t *pte, unsigned long addr,
 		   unsigned long next, struct mm_walk *walk)
 {
 	pte_val(*pte) &= ~_PAGE_CI;
@@ -67,9 +71,19 @@ static int page_clear_nocache(pte_t *pte, unsigned long addr,
  * cache-inhibit bit on those pages, and makes sure that the pages are
  * flushed out of the cache before they are used.
  *
+ * If the NON_CONSISTENT attribute is set, then this function just
+ * returns "normal", cachable memory.
+ *
+ * There are additional flags WEAK_ORDERING and WRITE_COMBINE to take
+ * into consideration here, too.  All current known implementations of
+ * the OR1K support only strongly ordered memory accesses, so that flag
+ * is being ignored for now; uncached but write-combined memory is a
+ * missing feature of the OR1K.
  */
-void *or1k_dma_alloc_coherent(struct device *dev, size_t size,
-			      dma_addr_t *dma_handle, gfp_t gfp)
+static void *
+or1k_dma_alloc(struct device *dev, size_t size,
+	       dma_addr_t *dma_handle, gfp_t gfp,
+	       struct dma_attrs *attrs)
 {
 	unsigned long va;
 	void *page;
@@ -87,6 +101,7 @@ void *or1k_dma_alloc_coherent(struct device *dev, size_t size,
 
 	va = (unsigned long)page;
 
+	if (!dma_get_attr(DMA_ATTR_NON_CONSISTENT, attrs)) {
 		/*
 		 * We need to iterate through the pages, clearing the dcache for
 		 * them and setting the cache-inhibit bit.
@@ -95,12 +110,14 @@ void *or1k_dma_alloc_coherent(struct device *dev, size_t size,
 			free_pages_exact(page, size);
 			return NULL;
 		}
+	}
 
 	return (void *)va;
 }
 
-void or1k_dma_free_coherent(struct device *dev, size_t size, void *vaddr,
-			    dma_addr_t dma_handle)
+static void
+or1k_dma_free(struct device *dev, size_t size, void *vaddr,
+	      dma_addr_t dma_handle, struct dma_attrs *attrs)
 {
 	unsigned long va = (unsigned long)vaddr;
 	struct mm_walk walk = {
@@ -108,13 +125,16 @@ void or1k_dma_free_coherent(struct device *dev, size_t size, void *vaddr,
 		.mm = &init_mm
 	};
 
+	if (!dma_get_attr(DMA_ATTR_NON_CONSISTENT, attrs)) {
 		/* walk_page_range shouldn't be able to fail here */
 		WARN_ON(walk_page_range(va, va + size, &walk));
+	}
 
 	free_pages_exact(vaddr, size);
 }
 
-dma_addr_t or1k_map_page(struct device *dev, struct page *page,
+static dma_addr_t
+or1k_map_page(struct device *dev, struct page *page,
 	      unsigned long offset, size_t size,
 	      enum dma_data_direction dir,
 	      struct dma_attrs *attrs)
@@ -147,14 +167,16 @@ dma_addr_t or1k_map_page(struct device *dev, struct page *page,
 	return addr;
 }
 
-void or1k_unmap_page(struct device *dev, dma_addr_t dma_handle,
+static void
+or1k_unmap_page(struct device *dev, dma_addr_t dma_handle,
 		size_t size, enum dma_data_direction dir,
 		struct dma_attrs *attrs)
 {
 	/* Nothing special to do here... */
 }
 
-int or1k_map_sg(struct device *dev, struct scatterlist *sg,
+static int
+or1k_map_sg(struct device *dev, struct scatterlist *sg,
 	    int nents, enum dma_data_direction dir,
 	    struct dma_attrs *attrs)
 {
@@ -169,7 +191,8 @@ int or1k_map_sg(struct device *dev, struct scatterlist *sg,
 	return nents;
 }
 
-void or1k_unmap_sg(struct device *dev, struct scatterlist *sg,
+static void
+or1k_unmap_sg(struct device *dev, struct scatterlist *sg,
 	      int nents, enum dma_data_direction dir,
 	      struct dma_attrs *attrs)
 {
@@ -181,7 +204,8 @@ void or1k_unmap_sg(struct device *dev, struct scatterlist *sg,
 	}
 }
 
-void or1k_sync_single_for_cpu(struct device *dev,
+static void
+or1k_sync_single_for_cpu(struct device *dev,
 			 dma_addr_t dma_handle, size_t size,
 			 enum dma_data_direction dir)
 {
@@ -193,7 +217,8 @@ void or1k_sync_single_for_cpu(struct device *dev,
 		mtspr(SPR_DCBIR, cl);
 }
 
-void or1k_sync_single_for_device(struct device *dev,
+static void
+or1k_sync_single_for_device(struct device *dev,
 			    dma_addr_t dma_handle, size_t size,
 			    enum dma_data_direction dir)
 {
@@ -205,6 +230,18 @@ void or1k_sync_single_for_device(struct device *dev,
 		mtspr(SPR_DCBFR, cl);
 }
 
+struct dma_map_ops or1k_dma_map_ops = {
+	.alloc = or1k_dma_alloc,
+	.free = or1k_dma_free,
+	.map_page = or1k_map_page,
+	.unmap_page = or1k_unmap_page,
+	.map_sg = or1k_map_sg,
+	.unmap_sg = or1k_unmap_sg,
+	.sync_single_for_cpu = or1k_sync_single_for_cpu,
+	.sync_single_for_device = or1k_sync_single_for_device,
+};
+EXPORT_SYMBOL(or1k_dma_map_ops);
+
 /* Number of entries preallocated for DMA-API debugging */
 #define PREALLOC_DMA_DEBUG_ENTRIES (1 << 16)
 

arch/openrisc/kernel/entry.S

@@ -1117,10 +1117,10 @@ ENTRY(sys_rt_sigreturn)
 ENTRY(sys_or1k_atomic)
 	/* FIXME: This ignores r3 and always does an XCHG */
 	DISABLE_INTERRUPTS(r17,r19)
-	l.lwz	r30,0(r4)
-	l.lwz	r28,0(r5)
-	l.sw	0(r4),r28
-	l.sw	0(r5),r30
+	l.lwz	r29,0(r4)
+	l.lwz	r27,0(r5)
+	l.sw	0(r4),r27
+	l.sw	0(r5),r29
 	ENABLE_INTERRUPTS(r17)
 	l.jr	r9
 	 l.or	r11,r0,r0

arch/openrisc/kernel/irq.c

@@ -14,17 +14,13 @@
  *      2 of the License, or (at your option) any later version.
  */
 
-#include <linux/ptrace.h>
-#include <linux/errno.h>
 #include <linux/interrupt.h>
 #include <linux/init.h>
 #include <linux/of.h>
 #include <linux/ftrace.h>
 #include <linux/irq.h>
-#include <linux/seq_file.h>
-#include <linux/kernel_stat.h>
 #include <linux/export.h>
-
+#include <linux/irqdomain.h>
 #include <linux/irqflags.h>
 
 /* read interrupt enabled status */
@@ -98,6 +94,7 @@ static void or1k_pic_mask_ack(struct irq_data *data)
 #endif
 }
 
+#if 0
 static int or1k_pic_set_type(struct irq_data *data, unsigned int flow_type)
 {
 	/* There's nothing to do in the PIC configuration when changing
@@ -107,43 +104,64 @@ static int or1k_pic_set_type(struct irq_data *data, unsigned int flow_type)
 
 	return irq_setup_alt_chip(data, flow_type);
 }
+#endif
+
+static struct irq_chip or1k_dev = {
+	.name = "or1k-PIC",
+	.irq_unmask = or1k_pic_unmask,
+	.irq_mask = or1k_pic_mask,
+	.irq_ack = or1k_pic_ack,
+	.irq_mask_ack = or1k_pic_mask_ack,
+};
+
+static struct irq_domain *root_domain;
 
 static inline int pic_get_irq(int first)
 {
-	int irq;
+	int hwirq;
 
-	irq = ffs(mfspr(SPR_PICSR) >> first);
+	hwirq = ffs(mfspr(SPR_PICSR) >> first);
+	if (!hwirq)
+		return NO_IRQ;
+	else
+		hwirq = hwirq + first -1;
 
-	return irq ? irq + first - 1 : NO_IRQ;
+	return irq_find_mapping(root_domain, hwirq);
 }
 
-static void __init or1k_irq_init(void)
-{
-	struct irq_chip_generic *gc;
-	struct irq_chip_type *ct;
 
-	/* Disable all interrupts until explicitly requested */
-	mtspr(SPR_PICMR, (0UL));
+static int or1k_map(struct irq_domain *d, unsigned int irq, irq_hw_number_t hw)
+{
+	irq_set_chip_and_handler_name(irq, &or1k_dev,
+				      handle_level_irq, "level");
+	irq_set_status_flags(irq, IRQ_LEVEL | IRQ_NOPROBE);
 
-	gc = irq_alloc_generic_chip("or1k-PIC", 1, 0, 0, handle_level_irq);
-	ct = gc->chip_types;
+	return 0;
+}
 
-	ct->chip.irq_unmask = or1k_pic_unmask;
-	ct->chip.irq_mask = or1k_pic_mask;
-	ct->chip.irq_ack = or1k_pic_ack;
-	ct->chip.irq_mask_ack = or1k_pic_mask_ack;
-	ct->chip.irq_set_type = or1k_pic_set_type;
+static const struct irq_domain_ops or1k_irq_domain_ops = {
+	.xlate = irq_domain_xlate_onecell,
+	.map = or1k_map,
+};
 
-	/* The OR1K PIC can handle both level and edge trigged
-	 * interrupts in roughly the same manner
+/*
+ * This sets up the IRQ domain for the PIC built in to the OpenRISC
+ * 1000 CPU.  This is the "root" domain as these are the interrupts
+ * that directly trigger an exception in the CPU.
  */
-#if 0
-	/* FIXME: chip.type??? */
-	ct->chip.type = IRQ_TYPE_EDGE_BOTH | IRQ_TYPE_LEVEL_MASK;
-#endif
+static void __init or1k_irq_init(void)
+{
+	struct device_node *intc = NULL;
+
+	/* The interrupt controller device node is mandatory */
+	intc = of_find_compatible_node(NULL, NULL, "opencores,or1k-pic");
+	BUG_ON(!intc);
 
-	irq_setup_generic_chip(gc, IRQ_MSK(NR_IRQS), 0,
-			       IRQ_NOREQUEST, IRQ_LEVEL | IRQ_NOPROBE);
+	/* Disable all interrupts until explicitly requested */
+	mtspr(SPR_PICMR, (0UL));
+
+	root_domain = irq_domain_add_linear(intc, 32,
+					    &or1k_irq_domain_ops, NULL);
 }
 
 void __init init_IRQ(void)
@@ -164,10 +182,3 @@ void __irq_entry do_IRQ(struct pt_regs *regs)
 	irq_exit();
 	set_irq_regs(old_regs);
 }
-
-unsigned int irq_create_of_mapping(struct device_node *controller,
-				   const u32 *intspec, unsigned int intsize)
-{
-	return intspec[0];
-}
-EXPORT_SYMBOL_GPL(irq_create_of_mapping);

arch/openrisc/mm/fault.c

@@ -54,6 +54,7 @@ asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long address,
 	struct vm_area_struct *vma;
 	siginfo_t info;
 	int fault;
+	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
 
 	tsk = current;
 
@@ -105,6 +106,7 @@ asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long address,
 	if (in_interrupt() || !mm)
 		goto no_context;
 
+retry:
 	down_read(&mm->mmap_sem);
 	vma = find_vma(mm, address);
 
@@ -143,6 +145,7 @@ asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long address,
 	if (write_acc) {
 		if (!(vma->vm_flags & VM_WRITE))
 			goto bad_area;
+		flags |= FAULT_FLAG_WRITE;
 	} else {
 		/* not present */
 		if (!(vma->vm_flags & (VM_READ | VM_EXEC)))
@@ -159,7 +162,11 @@ asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long address,
 	 * the fault.
 	 */
 
-	fault = handle_mm_fault(mm, vma, address, write_acc);
+	fault = handle_mm_fault(mm, vma, address, flags);
+
+	if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
+		return;
+
 	if (unlikely(fault & VM_FAULT_ERROR)) {
 		if (fault & VM_FAULT_OOM)
 			goto out_of_memory;
@@ -167,11 +174,24 @@ asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long address,
 			goto do_sigbus;
 		BUG();
 	}
+
+	if (flags & FAULT_FLAG_ALLOW_RETRY) {
 		/*RGD modeled on Cris */
 		if (fault & VM_FAULT_MAJOR)
 			tsk->maj_flt++;
 		else
 			tsk->min_flt++;
+		if (fault & VM_FAULT_RETRY) {
+			flags &= ~FAULT_FLAG_ALLOW_RETRY;
+
+			 /* No need to up_read(&mm->mmap_sem) as we would
+			 * have already released it in __lock_page_or_retry
+			 * in mm/filemap.c.
+			 */
+
+			goto retry;
+		}
+	}
 
 	up_read(&mm->mmap_sem);
 	return;