From c917a36f5fe551748eb37bd1efdcccc1045a02ab Mon Sep 17 00:00:00 2001
From: Vineet Gupta <vgupta@synopsys.com>
Date: Thu, 9 May 2013 13:45:12 +0530
Subject: ARC: [mm] serious bug in vaddr based icache flush

vaddr used to index the cache was clipped from the wrong end, and thus
would potentially fail to flush the correct lines.

The problem was dorment for so long because up until the recent
optimizations it was only used for ptrace break-point only flushes.

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>

diff --git a/arch/arc/mm/cache_arc700.c b/arch/arc/mm/cache_arc700.c
index c854cf9..d48afeb 100644
--- a/arch/arc/mm/cache_arc700.c
+++ b/arch/arc/mm/cache_arc700.c
@@ -421,7 +421,7 @@ static void __ic_line_inv_vaddr(unsigned long phy_start, unsigned long vaddr,
 	num_lines = DIV_ROUND_UP(sz, ARC_ICACHE_LINE_LEN);
 
 #if (CONFIG_ARC_MMU_VER > 2)
-	vaddr &= ~ICACHE_LINE_MASK;
+	vaddr &= ICACHE_LINE_MASK;
 	addr = phy_start;
 #else
 	/* bits 17:13 of vaddr go as bits 4:0 of paddr */
-- 
cgit v0.10.2


From a690984d60abcb627ce1bcc0300a14043b5e224a Mon Sep 17 00:00:00 2001
From: Vineet Gupta <vgupta@synopsys.com>
Date: Thu, 9 May 2013 14:00:51 +0530
Subject: ARC: [mm] refactor the core (i|d)cache line ops loops

Nothing semantical
* simplify the alignement code by using & operation only
* rename variables clearly as paddr

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>

diff --git a/arch/arc/mm/cache_arc700.c b/arch/arc/mm/cache_arc700.c
index d48afeb..8a760e2 100644
--- a/arch/arc/mm/cache_arc700.c
+++ b/arch/arc/mm/cache_arc700.c
@@ -270,21 +270,20 @@ static inline void __dc_entire_op(const int cacheop)
  * Doesn't deal with type-of-op/IRQ-disabling/waiting-for-flush-to-complete
  * It's sole purpose is to help gcc generate ZOL
  */
-static inline void __dc_line_loop(unsigned long start, unsigned long sz,
-					  int aux_reg)
+static inline void __dc_line_loop(unsigned long paddr, unsigned long sz,
+				  int aux_reg)
 {
-	int num_lines, slack;
+	int num_lines;
 
 	/* Ensure we properly floor/ceil the non-line aligned/sized requests
-	 * and have @start - aligned to cache line and integral @num_lines.
+	 * and have @paddr - aligned to cache line and integral @num_lines.
 	 * This however can be avoided for page sized since:
-	 *  -@start will be cache-line aligned already (being page aligned)
+	 *  -@paddr will be cache-line aligned already (being page aligned)
 	 *  -@sz will be integral multiple of line size (being page sized).
 	 */
 	if (!(__builtin_constant_p(sz) && sz == PAGE_SIZE)) {
-		slack = start & ~DCACHE_LINE_MASK;
-		sz += slack;
-		start -= slack;
+		sz += paddr & ~DCACHE_LINE_MASK;
+		paddr &= DCACHE_LINE_MASK;
 	}
 
 	num_lines = DIV_ROUND_UP(sz, ARC_DCACHE_LINE_LEN);
@@ -298,17 +297,17 @@ static inline void __dc_line_loop(unsigned long start, unsigned long sz,
 		 * doesn't support aliasing configs for D$, yet.
 		 * Thus paddr is enough to provide both tag and index.
 		 */
-		write_aux_reg(ARC_REG_DC_PTAG, start);
+		write_aux_reg(ARC_REG_DC_PTAG, paddr);
 #endif
-		write_aux_reg(aux_reg, start);
-		start += ARC_DCACHE_LINE_LEN;
+		write_aux_reg(aux_reg, paddr);
+		paddr += ARC_DCACHE_LINE_LEN;
 	}
 }
 
 /*
  * D-Cache : Per Line INV (discard or wback+discard) or FLUSH (wback)
  */
-static inline void __dc_line_op(unsigned long start, unsigned long sz,
+static inline void __dc_line_op(unsigned long paddr, unsigned long sz,
 					const int cacheop)
 {
 	unsigned long flags, tmp = tmp;
@@ -332,7 +331,7 @@ static inline void __dc_line_op(unsigned long start, unsigned long sz,
 	else
 		aux = ARC_REG_DC_FLDL;
 
-	__dc_line_loop(start, sz, aux);
+	__dc_line_loop(paddr, sz, aux);
 
 	if (cacheop & OP_FLUSH)	/* flush / flush-n-inv both wait */
 		wait_for_flush();
@@ -347,7 +346,7 @@ static inline void __dc_line_op(unsigned long start, unsigned long sz,
 #else
 
 #define __dc_entire_op(cacheop)
-#define __dc_line_op(start, sz, cacheop)
+#define __dc_line_op(paddr, sz, cacheop)
 
 #endif /* CONFIG_ARC_HAS_DCACHE */
 
@@ -399,49 +398,45 @@ static inline void __dc_line_op(unsigned long start, unsigned long sz,
 /***********************************************************
  * Machine specific helper for per line I-Cache invalidate.
  */
-static void __ic_line_inv_vaddr(unsigned long phy_start, unsigned long vaddr,
+static void __ic_line_inv_vaddr(unsigned long paddr, unsigned long vaddr,
 				unsigned long sz)
 {
 	unsigned long flags;
-	int num_lines, slack;
-	unsigned int addr;
+	int num_lines;
 
 	/*
 	 * Ensure we properly floor/ceil the non-line aligned/sized requests:
 	 * However page sized flushes can be compile time optimised.
-	 *  -@phy_start will be cache-line aligned already (being page aligned)
+	 *  -@paddr will be cache-line aligned already (being page aligned)
 	 *  -@sz will be integral multiple of line size (being page sized).
 	 */
 	if (!(__builtin_constant_p(sz) && sz == PAGE_SIZE)) {
-		slack = phy_start & ~ICACHE_LINE_MASK;
-		sz += slack;
-		phy_start -= slack;
+		sz += paddr & ~ICACHE_LINE_MASK;
+		paddr &= ICACHE_LINE_MASK;
+		vaddr &= ICACHE_LINE_MASK;
 	}
 
 	num_lines = DIV_ROUND_UP(sz, ARC_ICACHE_LINE_LEN);
 
-#if (CONFIG_ARC_MMU_VER > 2)
-	vaddr &= ICACHE_LINE_MASK;
-	addr = phy_start;
-#else
+#if (CONFIG_ARC_MMU_VER <= 2)
 	/* bits 17:13 of vaddr go as bits 4:0 of paddr */
-	addr = phy_start | ((vaddr >> 13) & 0x1F);
+	paddr |= (vaddr >> PAGE_SHIFT) & 0x1F;
 #endif
 
 	local_irq_save(flags);
 	while (num_lines-- > 0) {
 #if (CONFIG_ARC_MMU_VER > 2)
 		/* tag comes from phy addr */
-		write_aux_reg(ARC_REG_IC_PTAG, addr);
+		write_aux_reg(ARC_REG_IC_PTAG, paddr);
 
 		/* index bits come from vaddr */
 		write_aux_reg(ARC_REG_IC_IVIL, vaddr);
 		vaddr += ARC_ICACHE_LINE_LEN;
 #else
 		/* paddr contains stuffed vaddrs bits */
-		write_aux_reg(ARC_REG_IC_IVIL, addr);
+		write_aux_reg(ARC_REG_IC_IVIL, paddr);
 #endif
-		addr += ARC_ICACHE_LINE_LEN;
+		paddr += ARC_ICACHE_LINE_LEN;
 	}
 	local_irq_restore(flags);
 }
-- 
cgit v0.10.2


From 6ec18a81b22ab2b40df8424f2b5fc6be20ccad87 Mon Sep 17 00:00:00 2001
From: Vineet Gupta <vgupta@synopsys.com>
Date: Thu, 9 May 2013 15:10:18 +0530
Subject: ARC: [mm] Aliasing VIPT dcache support 1/4

This preps the low level dcache flush helpers to take vaddr argument in
addition to the existing paddr to properly flush the VIPT dcache

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>

diff --git a/arch/arc/include/asm/cacheflush.h b/arch/arc/include/asm/cacheflush.h
index ee1f6ea..ed820bcb 100644
--- a/arch/arc/include/asm/cacheflush.h
+++ b/arch/arc/include/asm/cacheflush.h
@@ -33,7 +33,7 @@ void flush_cache_all(void);
 void flush_icache_range(unsigned long start, unsigned long end);
 void __sync_icache_dcache(unsigned long paddr, unsigned long vaddr, int len);
 void __inv_icache_page(unsigned long paddr, unsigned long vaddr);
-void __flush_dcache_page(unsigned long paddr);
+void __flush_dcache_page(unsigned long paddr, unsigned long vaddr);
 
 #define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 1
 
diff --git a/arch/arc/mm/cache_arc700.c b/arch/arc/mm/cache_arc700.c
index 8a760e2..a9a3708 100644
--- a/arch/arc/mm/cache_arc700.c
+++ b/arch/arc/mm/cache_arc700.c
@@ -269,9 +269,10 @@ static inline void __dc_entire_op(const int cacheop)
  * Per Line Operation on D-Cache
  * Doesn't deal with type-of-op/IRQ-disabling/waiting-for-flush-to-complete
  * It's sole purpose is to help gcc generate ZOL
+ * (aliasing VIPT dcache flushing needs both vaddr and paddr)
  */
-static inline void __dc_line_loop(unsigned long paddr, unsigned long sz,
-				  int aux_reg)
+static inline void __dc_line_loop(unsigned long paddr, unsigned long vaddr,
+				  unsigned long sz, const int aux_reg)
 {
 	int num_lines;
 
@@ -284,31 +285,41 @@ static inline void __dc_line_loop(unsigned long paddr, unsigned long sz,
 	if (!(__builtin_constant_p(sz) && sz == PAGE_SIZE)) {
 		sz += paddr & ~DCACHE_LINE_MASK;
 		paddr &= DCACHE_LINE_MASK;
+		vaddr &= DCACHE_LINE_MASK;
 	}
 
 	num_lines = DIV_ROUND_UP(sz, ARC_DCACHE_LINE_LEN);
 
+#if (CONFIG_ARC_MMU_VER <= 2)
+	paddr |= (vaddr >> PAGE_SHIFT) & 0x1F;
+#endif
+
 	while (num_lines-- > 0) {
 #if (CONFIG_ARC_MMU_VER > 2)
 		/*
 		 * Just as for I$, in MMU v3, D$ ops also require
 		 * "tag" bits in DC_PTAG, "index" bits in FLDL,IVDL ops
-		 * But we pass phy addr for both. This works since Linux
-		 * doesn't support aliasing configs for D$, yet.
-		 * Thus paddr is enough to provide both tag and index.
 		 */
 		write_aux_reg(ARC_REG_DC_PTAG, paddr);
-#endif
+
+		write_aux_reg(aux_reg, vaddr);
+		vaddr += ARC_DCACHE_LINE_LEN;
+#else
+		/* paddr contains stuffed vaddrs bits */
 		write_aux_reg(aux_reg, paddr);
+#endif
 		paddr += ARC_DCACHE_LINE_LEN;
 	}
 }
 
+/* For kernel mappings cache op index is same as paddr */
+#define __dc_line_op_k(p, sz, op)	__dc_line_op(p, p, sz, op)
+
 /*
  * D-Cache : Per Line INV (discard or wback+discard) or FLUSH (wback)
  */
-static inline void __dc_line_op(unsigned long paddr, unsigned long sz,
-					const int cacheop)
+static inline void __dc_line_op(unsigned long paddr, unsigned long vaddr,
+				unsigned long sz, const int cacheop)
 {
 	unsigned long flags, tmp = tmp;
 	int aux;
@@ -331,7 +342,7 @@ static inline void __dc_line_op(unsigned long paddr, unsigned long sz,
 	else
 		aux = ARC_REG_DC_FLDL;
 
-	__dc_line_loop(paddr, sz, aux);
+	__dc_line_loop(paddr, vaddr, sz, aux);
 
 	if (cacheop & OP_FLUSH)	/* flush / flush-n-inv both wait */
 		wait_for_flush();
@@ -346,7 +357,8 @@ static inline void __dc_line_op(unsigned long paddr, unsigned long sz,
 #else
 
 #define __dc_entire_op(cacheop)
-#define __dc_line_op(paddr, sz, cacheop)
+#define __dc_line_op(paddr, vaddr, sz, cacheop)
+#define __dc_line_op_k(paddr, sz, cacheop)
 
 #endif /* CONFIG_ARC_HAS_DCACHE */
 
@@ -462,19 +474,19 @@ EXPORT_SYMBOL(flush_dcache_page);
 
 void dma_cache_wback_inv(unsigned long start, unsigned long sz)
 {
-	__dc_line_op(start, sz, OP_FLUSH_N_INV);
+	__dc_line_op_k(start, sz, OP_FLUSH_N_INV);
 }
 EXPORT_SYMBOL(dma_cache_wback_inv);
 
 void dma_cache_inv(unsigned long start, unsigned long sz)
 {
-	__dc_line_op(start, sz, OP_INV);
+	__dc_line_op_k(start, sz, OP_INV);
 }
 EXPORT_SYMBOL(dma_cache_inv);
 
 void dma_cache_wback(unsigned long start, unsigned long sz)
 {
-	__dc_line_op(start, sz, OP_FLUSH);
+	__dc_line_op_k(start, sz, OP_FLUSH);
 }
 EXPORT_SYMBOL(dma_cache_wback);
 
@@ -555,7 +567,7 @@ void __sync_icache_dcache(unsigned long paddr, unsigned long vaddr, int len)
 
 	local_irq_save(flags);
 	__ic_line_inv_vaddr(paddr, vaddr, len);
-	__dc_line_op(paddr, len, OP_FLUSH);
+	__dc_line_op(paddr, vaddr, len, OP_FLUSH);
 	local_irq_restore(flags);
 }
 
@@ -565,9 +577,13 @@ void __inv_icache_page(unsigned long paddr, unsigned long vaddr)
 	__ic_line_inv_vaddr(paddr, vaddr, PAGE_SIZE);
 }
 
-void __flush_dcache_page(unsigned long paddr)
+/*
+ * wrapper to clearout kernel or userspace mappings of a page
+ * For kernel mappings @vaddr == @paddr
+ */
+void __flush_dcache_page(unsigned long paddr, unsigned long vaddr)
 {
-	__dc_line_op(paddr, PAGE_SIZE, OP_FLUSH_N_INV);
+	__dc_line_op(paddr, vaddr & PAGE_MASK, PAGE_SIZE, OP_FLUSH_N_INV);
 }
 
 void flush_icache_all(void)
diff --git a/arch/arc/mm/tlb.c b/arch/arc/mm/tlb.c
index 003d69a..5810c7a 100644
--- a/arch/arc/mm/tlb.c
+++ b/arch/arc/mm/tlb.c
@@ -438,7 +438,7 @@ void update_mmu_cache(struct vm_area_struct *vma, unsigned long vaddr_unaligned,
 		int dirty = test_and_clear_bit(PG_arch_1, &page->flags);
 		if (dirty) {
 			unsigned long paddr =  pte_val(*ptep) & PAGE_MASK;
-			__flush_dcache_page(paddr);
+			__flush_dcache_page(paddr, paddr);
 			__inv_icache_page(paddr, vaddr);
 		}
 	}
-- 
cgit v0.10.2


From 4102b53392d6397d80b6e09b516517efacf7ea77 Mon Sep 17 00:00:00 2001
From: Vineet Gupta <vgupta@synopsys.com>
Date: Thu, 9 May 2013 21:54:51 +0530
Subject: ARC: [mm] Aliasing VIPT dcache support 2/4

This is the meat of the series which prevents any dcache alias creation
by always keeping the U and K mapping of a page congruent.
If a mapping already exists, and other tries to access the page, prev
one is flushed to physical page (wback+inv)

Essentially flush_dcache_page()/copy_user_highpage() create K-mapping
of a page, but try to defer flushing, unless U-mapping exist.
When page is actually mapped to userspace, update_mmu_cache() flushes
the K-mapping (in certain cases this can be optimised out)

Additonally flush_cache_mm(), flush_cache_range(), flush_cache_page()
handle the puring of stale userspace mappings on exit/munmap...

flush_anon_page() handles the existing U-mapping for anon page before
kernel reads it via the GUP path.

Note that while not complete, this is enough to boot a simple
dynamically linked Busybox based rootfs

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>

diff --git a/arch/arc/Kconfig b/arch/arc/Kconfig
index 491ae79..5917099 100644
--- a/arch/arc/Kconfig
+++ b/arch/arc/Kconfig
@@ -182,6 +182,10 @@ config ARC_CACHE_PAGES
 	  Note that Global I/D ENABLE + Per Page DISABLE works but corollary
 	  Global DISABLE + Per Page ENABLE won't work
 
+config ARC_CACHE_VIPT_ALIASING
+	bool "Support VIPT Aliasing D$"
+	default n
+
 endif	#ARC_CACHE
 
 config ARC_HAS_ICCM
diff --git a/arch/arc/include/asm/cacheflush.h b/arch/arc/include/asm/cacheflush.h
index ed820bcb..d692fbb 100644
--- a/arch/arc/include/asm/cacheflush.h
+++ b/arch/arc/include/asm/cacheflush.h
@@ -50,18 +50,55 @@ void dma_cache_wback(unsigned long start, unsigned long sz);
 #define flush_cache_vmap(start, end)		flush_cache_all()
 #define flush_cache_vunmap(start, end)		flush_cache_all()
 
-/*
- * VM callbacks when entire/range of user-space V-P mappings are
- * torn-down/get-invalidated
- *
- * Currently we don't support D$ aliasing configs for our VIPT caches
- * NOPS for VIPT Cache with non-aliasing D$ configurations only
- */
-#define flush_cache_dup_mm(mm)			/* called on fork */
+#define flush_cache_dup_mm(mm)			/* called on fork (VIVT only) */
+
+#ifndef CONFIG_ARC_CACHE_VIPT_ALIASING
+
 #define flush_cache_mm(mm)			/* called on munmap/exit */
 #define flush_cache_range(mm, u_vstart, u_vend)
 #define flush_cache_page(vma, u_vaddr, pfn)	/* PF handling/COW-break */
 
+#else	/* VIPT aliasing dcache */
+
+/* To clear out stale userspace mappings */
+void flush_cache_mm(struct mm_struct *mm);
+void flush_cache_range(struct vm_area_struct *vma,
+	unsigned long start,unsigned long end);
+void flush_cache_page(struct vm_area_struct *vma,
+	unsigned long user_addr, unsigned long page);
+
+/*
+ * To make sure that userspace mapping is flushed to memory before
+ * get_user_pages() uses a kernel mapping to access the page
+ */
+#define ARCH_HAS_FLUSH_ANON_PAGE
+void flush_anon_page(struct vm_area_struct *vma,
+	struct page *page, unsigned long u_vaddr);
+
+#endif	/* CONFIG_ARC_CACHE_VIPT_ALIASING */
+
+/*
+ * Simple wrapper over config option
+ * Bootup code ensures that hardware matches kernel configuration
+ */
+static inline int cache_is_vipt_aliasing(void)
+{
+#ifdef CONFIG_ARC_CACHE_VIPT_ALIASING
+	return 1;
+#else
+	return 0;
+#endif
+}
+
+#define CACHE_COLOR(addr)	(((unsigned long)(addr) >> (PAGE_SHIFT)) & 3)
+
+/*
+ * checks if two addresses (after page aligning) index into same cache set
+ */
+#define addr_not_cache_congruent(addr1, addr2)				\
+	cache_is_vipt_aliasing() ? 					\
+		(CACHE_COLOR(addr1) != CACHE_COLOR(addr2)) : 0		\
+
 #define copy_to_user_page(vma, page, vaddr, dst, src, len)		\
 do {									\
 	memcpy(dst, src, len);						\
diff --git a/arch/arc/include/asm/page.h b/arch/arc/include/asm/page.h
index bdf5461..374a355 100644
--- a/arch/arc/include/asm/page.h
+++ b/arch/arc/include/asm/page.h
@@ -16,13 +16,27 @@
 #define get_user_page(vaddr)		__get_free_page(GFP_KERNEL)
 #define free_user_page(page, addr)	free_page(addr)
 
-/* TBD: for now don't worry about VIPT D$ aliasing */
 #define clear_page(paddr)		memset((paddr), 0, PAGE_SIZE)
 #define copy_page(to, from)		memcpy((to), (from), PAGE_SIZE)
 
+#ifndef CONFIG_ARC_CACHE_VIPT_ALIASING
+
 #define clear_user_page(addr, vaddr, pg)	clear_page(addr)
 #define copy_user_page(vto, vfrom, vaddr, pg)	copy_page(vto, vfrom)
 
+#else	/* VIPT aliasing dcache */
+
+struct vm_area_struct;
+struct page;
+
+#define __HAVE_ARCH_COPY_USER_HIGHPAGE
+
+void copy_user_highpage(struct page *to, struct page *from,
+			unsigned long u_vaddr, struct vm_area_struct *vma);
+void clear_user_page(void *to, unsigned long u_vaddr, struct page *page);
+
+#endif	/* CONFIG_ARC_CACHE_VIPT_ALIASING */
+
 #undef STRICT_MM_TYPECHECKS
 
 #ifdef STRICT_MM_TYPECHECKS
diff --git a/arch/arc/include/asm/tlb.h b/arch/arc/include/asm/tlb.h
index fe91719..85b6df8 100644
--- a/arch/arc/include/asm/tlb.h
+++ b/arch/arc/include/asm/tlb.h
@@ -30,13 +30,20 @@ do {						\
 /*
  * This pair is called at time of munmap/exit to flush cache and TLB entries
  * for mappings being torn down.
- * 1) cache-flush part -implemented via tlb_start_vma( ) can be NOP (for now)
- *    as we don't support aliasing configs in our VIPT D$.
+ * 1) cache-flush part -implemented via tlb_start_vma( ) for VIPT aliasing D$
  * 2) tlb-flush part - implemted via tlb_end_vma( ) flushes the TLB range
  *
  * Note, read http://lkml.org/lkml/2004/1/15/6
  */
+#ifndef CONFIG_ARC_CACHE_VIPT_ALIASING
 #define tlb_start_vma(tlb, vma)
+#else
+#define tlb_start_vma(tlb, vma)						\
+do {									\
+	if (!tlb->fullmm)						\
+		flush_cache_range(vma, vma->vm_start, vma->vm_end);	\
+} while(0)
+#endif
 
 #define tlb_end_vma(tlb, vma)						\
 do {									\
diff --git a/arch/arc/mm/cache_arc700.c b/arch/arc/mm/cache_arc700.c
index a9a3708..9887195 100644
--- a/arch/arc/mm/cache_arc700.c
+++ b/arch/arc/mm/cache_arc700.c
@@ -68,6 +68,7 @@
 #include <linux/mmu_context.h>
 #include <linux/syscalls.h>
 #include <linux/uaccess.h>
+#include <linux/pagemap.h>
 #include <asm/cacheflush.h>
 #include <asm/cachectl.h>
 #include <asm/setup.h>
@@ -138,6 +139,7 @@ void __cpuinit arc_cache_init(void)
 	struct cpuinfo_arc_cache *ic = &cpuinfo_arc700[cpu].icache;
 	struct cpuinfo_arc_cache *dc = &cpuinfo_arc700[cpu].dcache;
 	int way_pg_ratio = way_pg_ratio;
+	int dcache_does_alias;
 	char str[256];
 
 	printk(arc_cache_mumbojumbo(0, str, sizeof(str)));
@@ -184,9 +186,13 @@ chk_dc:
 		panic("Cache H/W doesn't match kernel Config");
 	}
 
+	dcache_does_alias = (dc->sz / ARC_DCACHE_WAYS) > PAGE_SIZE;
+
 	/* check for D-Cache aliasing */
-	if ((dc->sz / ARC_DCACHE_WAYS) > PAGE_SIZE)
-		panic("D$ aliasing not handled right now\n");
+	if (dcache_does_alias && !cache_is_vipt_aliasing())
+		panic("Enable CONFIG_ARC_CACHE_VIPT_ALIASING\n");
+	else if (!dcache_does_alias && cache_is_vipt_aliasing())
+		panic("Don't need CONFIG_ARC_CACHE_VIPT_ALIASING\n");
 #endif
 
 	/* Set the default Invalidate Mode to "simpy discard dirty lines"
@@ -312,7 +318,7 @@ static inline void __dc_line_loop(unsigned long paddr, unsigned long vaddr,
 	}
 }
 
-/* For kernel mappings cache op index is same as paddr */
+/* For kernel mappings cache operation: index is same as paddr */
 #define __dc_line_op_k(p, sz, op)	__dc_line_op(p, p, sz, op)
 
 /*
@@ -464,10 +470,47 @@ static void __ic_line_inv_vaddr(unsigned long paddr, unsigned long vaddr,
  * Exported APIs
  */
 
+/*
+ * Handle cache congruency of kernel and userspace mappings of page when kernel
+ * writes-to/reads-from
+ *
+ * The idea is to defer flushing of kernel mapping after a WRITE, possible if:
+ *  -dcache is NOT aliasing, hence any U/K-mappings of page are congruent
+ *  -U-mapping doesn't exist yet for page (finalised in update_mmu_cache)
+ *  -In SMP, if hardware caches are coherent
+ *
+ * There's a corollary case, where kernel READs from a userspace mapped page.
+ * If the U-mapping is not congruent to to K-mapping, former needs flushing.
+ */
 void flush_dcache_page(struct page *page)
 {
-	/* Make a note that dcache is not yet flushed for this page */
-	set_bit(PG_arch_1, &page->flags);
+	struct address_space *mapping;
+
+	if (!cache_is_vipt_aliasing()) {
+		set_bit(PG_arch_1, &page->flags);
+		return;
+	}
+
+	/* don't handle anon pages here */
+	mapping = page_mapping(page);
+	if (!mapping)
+		return;
+
+	/*
+	 * pagecache page, file not yet mapped to userspace
+	 * Make a note that K-mapping is dirty
+	 */
+	if (!mapping_mapped(mapping)) {
+		set_bit(PG_arch_1, &page->flags);
+	} else if (page_mapped(page)) {
+
+		/* kernel reading from page with U-mapping */
+		void *paddr = page_address(page);
+		unsigned long vaddr = page->index << PAGE_CACHE_SHIFT;
+
+		if (addr_not_cache_congruent(paddr, vaddr))
+			__flush_dcache_page(paddr, vaddr);
+	}
 }
 EXPORT_SYMBOL(flush_dcache_page);
 
@@ -612,6 +655,87 @@ noinline void flush_cache_all(void)
 
 }
 
+#ifdef CONFIG_ARC_CACHE_VIPT_ALIASING
+
+void flush_cache_mm(struct mm_struct *mm)
+{
+	flush_cache_all();
+}
+
+void flush_cache_page(struct vm_area_struct *vma, unsigned long u_vaddr,
+		      unsigned long pfn)
+{
+	unsigned int paddr = pfn << PAGE_SHIFT;
+
+	__sync_icache_dcache(paddr, u_vaddr, PAGE_SIZE);
+}
+
+void flush_cache_range(struct vm_area_struct *vma, unsigned long start,
+		       unsigned long end)
+{
+	flush_cache_all();
+}
+
+void copy_user_highpage(struct page *to, struct page *from,
+	unsigned long u_vaddr, struct vm_area_struct *vma)
+{
+	void *kfrom = page_address(from);
+	void *kto = page_address(to);
+	int clean_src_k_mappings = 0;
+
+	/*
+	 * If SRC page was already mapped in userspace AND it's U-mapping is
+	 * not congruent with K-mapping, sync former to physical page so that
+	 * K-mapping in memcpy below, sees the right data
+	 *
+	 * Note that while @u_vaddr refers to DST page's userspace vaddr, it is
+	 * equally valid for SRC page as well
+	 */
+	if (page_mapped(from) && addr_not_cache_congruent(kfrom, u_vaddr)) {
+		__flush_dcache_page(kfrom, u_vaddr);
+		clean_src_k_mappings = 1;
+	}
+
+	copy_page(kto, kfrom);
+
+	/*
+	 * Mark DST page K-mapping as dirty for a later finalization by
+	 * update_mmu_cache(). Although the finalization could have been done
+	 * here as well (given that both vaddr/paddr are available).
+	 * But update_mmu_cache() already has code to do that for other
+	 * non copied user pages (e.g. read faults which wire in pagecache page
+	 * directly).
+	 */
+	set_bit(PG_arch_1, &to->flags);
+
+	/*
+	 * if SRC was already usermapped and non-congruent to kernel mapping
+	 * sync the kernel mapping back to physical page
+	 */
+	if (clean_src_k_mappings) {
+		__flush_dcache_page(kfrom, kfrom);
+	} else {
+		set_bit(PG_arch_1, &from->flags);
+	}
+}
+
+void clear_user_page(void *to, unsigned long u_vaddr, struct page *page)
+{
+	clear_page(to);
+	set_bit(PG_arch_1, &page->flags);
+}
+
+void flush_anon_page(struct vm_area_struct *vma, struct page *page,
+		     unsigned long u_vaddr)
+{
+	/* TBD: do we really need to clear the kernel mapping */
+	__flush_dcache_page(page_address(page), u_vaddr);
+	__flush_dcache_page(page_address(page), page_address(page));
+
+}
+
+#endif
+
 /**********************************************************************
  * Explicit Cache flush request from user space via syscall
  * Needed for JITs which generate code on the fly
diff --git a/arch/arc/mm/tlb.c b/arch/arc/mm/tlb.c
index 5810c7a..066145b 100644
--- a/arch/arc/mm/tlb.c
+++ b/arch/arc/mm/tlb.c
@@ -421,25 +421,40 @@ void create_tlb(struct vm_area_struct *vma, unsigned long address, pte_t *ptep)
 /*
  * Called at the end of pagefault, for a userspace mapped page
  *  -pre-install the corresponding TLB entry into MMU
- *  -Finalize the delayed D-cache flush (wback+inv kernel mapping)
+ *  -Finalize the delayed D-cache flush of kernel mapping of page due to
+ *  	flush_dcache_page(), copy_user_page()
+ *
+ * Note that flush (when done) involves both WBACK - so physical page is
+ * in sync as well as INV - so any non-congruent aliases don't remain
  */
 void update_mmu_cache(struct vm_area_struct *vma, unsigned long vaddr_unaligned,
 		      pte_t *ptep)
 {
 	unsigned long vaddr = vaddr_unaligned & PAGE_MASK;
+	unsigned long paddr = pte_val(*ptep) & PAGE_MASK;
 
 	create_tlb(vma, vaddr, ptep);
 
-	/* icache doesn't snoop dcache, thus needs to be made coherent here */
-	if (vma->vm_flags & VM_EXEC) {
+	/*
+	 * Exec page : Independent of aliasing/page-color considerations,
+	 *	       since icache doesn't snoop dcache on ARC, any dirty
+	 *	       K-mapping of a code page needs to be wback+inv so that
+	 *	       icache fetch by userspace sees code correctly.
+	 * !EXEC page: If K-mapping is NOT congruent to U-mapping, flush it
+	 *	       so userspace sees the right data.
+	 *  (Avoids the flush for Non-exec + congruent mapping case)
+	 */
+	if (vma->vm_flags & VM_EXEC || addr_not_cache_congruent(paddr, vaddr)) {
 		struct page *page = pfn_to_page(pte_pfn(*ptep));
 
-		/* if page was dcache dirty, flush now */
 		int dirty = test_and_clear_bit(PG_arch_1, &page->flags);
 		if (dirty) {
-			unsigned long paddr =  pte_val(*ptep) & PAGE_MASK;
+			/* wback + inv dcache lines */
 			__flush_dcache_page(paddr, paddr);
-			__inv_icache_page(paddr, vaddr);
+
+			/* invalidate any existing icache lines */
+			if (vma->vm_flags & VM_EXEC)
+				__inv_icache_page(paddr, vaddr);
 		}
 	}
 }
-- 
cgit v0.10.2


From de2a852cc0d4c4d6a9c22a597c9cc231f2e6ceb4 Mon Sep 17 00:00:00 2001
From: Vineet Gupta <vgupta@synopsys.com>
Date: Thu, 9 May 2013 21:55:27 +0530
Subject: ARC: [mm] Aliasing VIPT dcache support 3/4

Fix the one zillion warnings

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>

diff --git a/arch/arc/include/asm/cacheflush.h b/arch/arc/include/asm/cacheflush.h
index d692fbb..14a0fea 100644
--- a/arch/arc/include/asm/cacheflush.h
+++ b/arch/arc/include/asm/cacheflush.h
@@ -33,7 +33,9 @@ void flush_cache_all(void);
 void flush_icache_range(unsigned long start, unsigned long end);
 void __sync_icache_dcache(unsigned long paddr, unsigned long vaddr, int len);
 void __inv_icache_page(unsigned long paddr, unsigned long vaddr);
-void __flush_dcache_page(unsigned long paddr, unsigned long vaddr);
+void ___flush_dcache_page(unsigned long paddr, unsigned long vaddr);
+#define __flush_dcache_page(p, v)	\
+		___flush_dcache_page((unsigned long)p, (unsigned long)v)
 
 #define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 1
 
diff --git a/arch/arc/mm/cache_arc700.c b/arch/arc/mm/cache_arc700.c
index 9887195..2f12bca 100644
--- a/arch/arc/mm/cache_arc700.c
+++ b/arch/arc/mm/cache_arc700.c
@@ -624,7 +624,7 @@ void __inv_icache_page(unsigned long paddr, unsigned long vaddr)
  * wrapper to clearout kernel or userspace mappings of a page
  * For kernel mappings @vaddr == @paddr
  */
-void __flush_dcache_page(unsigned long paddr, unsigned long vaddr)
+void ___flush_dcache_page(unsigned long paddr, unsigned long vaddr)
 {
 	__dc_line_op(paddr, vaddr & PAGE_MASK, PAGE_SIZE, OP_FLUSH_N_INV);
 }
-- 
cgit v0.10.2


From 5bba49f5397c012d873c73860ad7b50c526e613b Mon Sep 17 00:00:00 2001
From: Vineet Gupta <vgupta@synopsys.com>
Date: Thu, 9 May 2013 19:20:43 +0530
Subject: ARC: [mm] Aliasing VIPT dcache support 4/4

Enforce congruency of userspace shared mappings

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>

diff --git a/arch/arc/include/asm/Kbuild b/arch/arc/include/asm/Kbuild
index 48af742..d8dd660 100644
--- a/arch/arc/include/asm/Kbuild
+++ b/arch/arc/include/asm/Kbuild
@@ -32,7 +32,6 @@ generic-y += resource.h
 generic-y += scatterlist.h
 generic-y += sembuf.h
 generic-y += shmbuf.h
-generic-y += shmparam.h
 generic-y += siginfo.h
 generic-y += socket.h
 generic-y += sockios.h
diff --git a/arch/arc/include/asm/cache.h b/arch/arc/include/asm/cache.h
index 6632273..d5555fe 100644
--- a/arch/arc/include/asm/cache.h
+++ b/arch/arc/include/asm/cache.h
@@ -55,9 +55,6 @@
 	: "r"(data), "r"(ptr));		\
 })
 
-/* used to give SHMLBA a value to avoid Cache Aliasing */
-extern unsigned int ARC_shmlba;
-
 #define ARCH_DMA_MINALIGN      L1_CACHE_BYTES
 
 /*
diff --git a/arch/arc/include/asm/cacheflush.h b/arch/arc/include/asm/cacheflush.h
index 14a0fea..9f841af 100644
--- a/arch/arc/include/asm/cacheflush.h
+++ b/arch/arc/include/asm/cacheflush.h
@@ -19,6 +19,7 @@
 #define _ASM_CACHEFLUSH_H
 
 #include <linux/mm.h>
+#include <asm/shmparam.h>
 
 /*
  * Semantically we need this because icache doesn't snoop dcache/dma.
diff --git a/arch/arc/include/asm/pgtable.h b/arch/arc/include/asm/pgtable.h
index b7e3668..1cc4720 100644
--- a/arch/arc/include/asm/pgtable.h
+++ b/arch/arc/include/asm/pgtable.h
@@ -395,6 +395,9 @@ void update_mmu_cache(struct vm_area_struct *vma, unsigned long address,
 
 #include <asm-generic/pgtable.h>
 
+/* to cope with aliasing VIPT cache */
+#define HAVE_ARCH_UNMAPPED_AREA
+
 /*
  * No page table caches to initialise
  */
diff --git a/arch/arc/include/asm/shmparam.h b/arch/arc/include/asm/shmparam.h
new file mode 100644
index 0000000..fffeecc
--- /dev/null
+++ b/arch/arc/include/asm/shmparam.h
@@ -0,0 +1,18 @@
+/*
+ * Copyright (C) 2013 Synopsys, Inc. (www.synopsys.com)
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef __ARC_ASM_SHMPARAM_H
+#define __ARC_ASM_SHMPARAM_H
+
+/* Handle upto 2 cache bins */
+#define	SHMLBA	(2 * PAGE_SIZE)
+
+/* Enforce SHMLBA in shmat */
+#define __ARCH_FORCE_SHMLBA
+
+#endif
diff --git a/arch/arc/mm/Makefile b/arch/arc/mm/Makefile
index 168dc14..ac95cc2 100644
--- a/arch/arc/mm/Makefile
+++ b/arch/arc/mm/Makefile
@@ -7,4 +7,4 @@
 #
 
 obj-y	:= extable.o ioremap.o dma.o fault.o init.o
-obj-y	+= tlb.o tlbex.o cache_arc700.o
+obj-y	+= tlb.o tlbex.o cache_arc700.o mmap.o
diff --git a/arch/arc/mm/mmap.c b/arch/arc/mm/mmap.c
new file mode 100644
index 0000000..2e06d56
--- /dev/null
+++ b/arch/arc/mm/mmap.c
@@ -0,0 +1,78 @@
+/*
+ * ARC700 mmap
+ *
+ * (started from arm version - for VIPT alias handling)
+ *
+ * Copyright (C) 2013 Synopsys, Inc. (www.synopsys.com)
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/mman.h>
+#include <linux/sched.h>
+#include <asm/cacheflush.h>
+
+#define COLOUR_ALIGN(addr, pgoff)			\
+	((((addr) + SHMLBA - 1) & ~(SHMLBA - 1)) +	\
+	 (((pgoff) << PAGE_SHIFT) & (SHMLBA - 1)))
+
+/*
+ * Ensure that shared mappings are correctly aligned to
+ * avoid aliasing issues with VIPT caches.
+ * We need to ensure that
+ * a specific page of an object is always mapped at a multiple of
+ * SHMLBA bytes.
+ */
+unsigned long
+arch_get_unmapped_area(struct file *filp, unsigned long addr,
+		unsigned long len, unsigned long pgoff, unsigned long flags)
+{
+	struct mm_struct *mm = current->mm;
+	struct vm_area_struct *vma;
+	int do_align = 0;
+	int aliasing = cache_is_vipt_aliasing();
+	struct vm_unmapped_area_info info;
+
+	/*
+	 * We only need to do colour alignment if D cache aliases.
+	 */
+	if (aliasing)
+		do_align = filp || (flags & MAP_SHARED);
+
+	/*
+	 * We enforce the MAP_FIXED case.
+	 */
+	if (flags & MAP_FIXED) {
+		if (aliasing && flags & MAP_SHARED &&
+		    (addr - (pgoff << PAGE_SHIFT)) & (SHMLBA - 1))
+			return -EINVAL;
+		return addr;
+	}
+
+	if (len > TASK_SIZE)
+		return -ENOMEM;
+
+	if (addr) {
+		if (do_align)
+			addr = COLOUR_ALIGN(addr, pgoff);
+		else
+			addr = PAGE_ALIGN(addr);
+
+		vma = find_vma(mm, addr);
+		if (TASK_SIZE - len >= addr &&
+		    (!vma || addr + len <= vma->vm_start))
+			return addr;
+	}
+
+	info.flags = 0;
+	info.length = len;
+	info.low_limit = mm->mmap_base;
+	info.high_limit = TASK_SIZE;
+	info.align_mask = do_align ? (PAGE_MASK & (SHMLBA - 1)) : 0;
+	info.align_offset = pgoff << PAGE_SHIFT;
+	return vm_unmapped_area(&info);
+}
-- 
cgit v0.10.2


From e7d5bab5efb94a82d80cab3ad68c357ce73ea596 Mon Sep 17 00:00:00 2001
From: Vineet Gupta <Vineet.Gupta1@synopsys.com>
Date: Wed, 17 Apr 2013 17:37:31 +0530
Subject: ARC: [TB10x] Remove GENERIC_GPIO

This tracks Alexandre Courbot's mainline GPIO rework

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Acked-by: Alexandre Courbot <acourbot@nvidia.com>

diff --git a/arch/arc/plat-tb10x/Kconfig b/arch/arc/plat-tb10x/Kconfig
index 4e12127..1d34521 100644
--- a/arch/arc/plat-tb10x/Kconfig
+++ b/arch/arc/plat-tb10x/Kconfig
@@ -27,10 +27,3 @@ menuconfig ARC_PLAT_TB10X
 	  Abilis Systems. TB10x is based on the ARC700 CPU architecture.
 	  Say Y if you are building a kernel for one of the SOCs in this
 	  series (e.g. TB100 or TB101). If in doubt say N.
-
-if ARC_PLAT_TB10X
-
-config GENERIC_GPIO
-	def_bool y
-
-endif
-- 
cgit v0.10.2