arm64: Cache maintenance routines

The patch adds functionality required for cache maintenance. The AArch64
architecture mandates non-aliasing VIPT or PIPT D-cache and VIPT (may
have aliases) or ASID-tagged VIVT I-cache. Cache maintenance operations
are automatically broadcast in hardware between CPUs.

Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Tony Lindgren <tony@atomide.com>
Acked-by: Nicolas Pitre <nico@linaro.org>
Acked-by: Olof Johansson <olof@lixom.net>
Acked-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>

arch/arm64/include/asm/cache.h

@@ -0,0 +1,32 @@
+/*
+ * Copyright (C) 2012 ARM Ltd.
+ *
+ * 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.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+#ifndef __ASM_CACHE_H
+#define __ASM_CACHE_H
+
+#define L1_CACHE_SHIFT		6
+#define L1_CACHE_BYTES		(1 << L1_CACHE_SHIFT)
+
+/*
+ * Memory returned by kmalloc() may be used for DMA, so we must make
+ * sure that all such allocations are cache aligned. Otherwise,
+ * unrelated code may cause parts of the buffer to be read into the
+ * cache before the transfer is done, causing old data to be seen by
+ * the CPU.
+ */
+#define ARCH_DMA_MINALIGN	L1_CACHE_BYTES
+#define ARCH_SLAB_MINALIGN	8
+
+#endif

arch/arm64/include/asm/cacheflush.h

@@ -0,0 +1,148 @@
+/*
+ * Based on arch/arm/include/asm/cacheflush.h
+ *
+ * Copyright (C) 1999-2002 Russell King.
+ * Copyright (C) 2012 ARM Ltd.
+ *
+ * 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.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+#ifndef __ASM_CACHEFLUSH_H
+#define __ASM_CACHEFLUSH_H
+
+#include <linux/mm.h>
+
+/*
+ * This flag is used to indicate that the page pointed to by a pte is clean
+ * and does not require cleaning before returning it to the user.
+ */
+#define PG_dcache_clean PG_arch_1
+
+/*
+ *	MM Cache Management
+ *	===================
+ *
+ *	The arch/arm64/mm/cache.S implements these methods.
+ *
+ *	Start addresses are inclusive and end addresses are exclusive; start
+ *	addresses should be rounded down, end addresses up.
+ *
+ *	See Documentation/cachetlb.txt for more information. Please note that
+ *	the implementation assumes non-aliasing VIPT D-cache and (aliasing)
+ *	VIPT or ASID-tagged VIVT I-cache.
+ *
+ *	flush_cache_all()
+ *
+ *		Unconditionally clean and invalidate the entire cache.
+ *
+ *	flush_cache_mm(mm)
+ *
+ *		Clean and invalidate all user space cache entries
+ *		before a change of page tables.
+ *
+ *	flush_icache_range(start, end)
+ *
+ *		Ensure coherency between the I-cache and the D-cache in the
+ *		region described by start, end.
+ *		- start  - virtual start address
+ *		- end    - virtual end address
+ *
+ *	__flush_cache_user_range(start, end)
+ *
+ *		Ensure coherency between the I-cache and the D-cache in the
+ *		region described by start, end.
+ *		- start  - virtual start address
+ *		- end    - virtual end address
+ *
+ *	__flush_dcache_area(kaddr, size)
+ *
+ *		Ensure that the data held in page is written back.
+ *		- kaddr  - page address
+ *		- size   - region size
+ */
+extern void flush_cache_all(void);
+extern void flush_cache_mm(struct mm_struct *mm);
+extern void flush_cache_range(struct vm_area_struct *vma, unsigned long start, unsigned long end);
+extern void flush_cache_page(struct vm_area_struct *vma, unsigned long user_addr, unsigned long pfn);
+extern void flush_icache_range(unsigned long start, unsigned long end);
+extern void __flush_dcache_area(void *addr, size_t len);
+extern void __flush_cache_user_range(unsigned long start, unsigned long end);
+
+/*
+ * Copy user data from/to a page which is mapped into a different
+ * processes address space.  Really, we want to allow our "user
+ * space" model to handle this.
+ */
+extern void copy_to_user_page(struct vm_area_struct *, struct page *,
+	unsigned long, void *, const void *, unsigned long);
+#define copy_from_user_page(vma, page, vaddr, dst, src, len) \
+	do {							\
+		memcpy(dst, src, len);				\
+	} while (0)
+
+#define flush_cache_dup_mm(mm) flush_cache_mm(mm)
+
+/*
+ * flush_dcache_page is used when the kernel has written to the page
+ * cache page at virtual address page->virtual.
+ *
+ * If this page isn't mapped (ie, page_mapping == NULL), or it might
+ * have userspace mappings, then we _must_ always clean + invalidate
+ * the dcache entries associated with the kernel mapping.
+ *
+ * Otherwise we can defer the operation, and clean the cache when we are
+ * about to change to user space.  This is the same method as used on SPARC64.
+ * See update_mmu_cache for the user space part.
+ */
+#define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 1
+extern void flush_dcache_page(struct page *);
+
+static inline void __flush_icache_all(void)
+{
+	asm("ic	ialluis");
+}
+
+#define flush_dcache_mmap_lock(mapping) \
+	spin_lock_irq(&(mapping)->tree_lock)
+#define flush_dcache_mmap_unlock(mapping) \
+	spin_unlock_irq(&(mapping)->tree_lock)
+
+#define flush_icache_user_range(vma,page,addr,len) \
+	flush_dcache_page(page)
+
+/*
+ * We don't appear to need to do anything here.  In fact, if we did, we'd
+ * duplicate cache flushing elsewhere performed by flush_dcache_page().
+ */
+#define flush_icache_page(vma,page)	do { } while (0)
+
+/*
+ * flush_cache_vmap() is used when creating mappings (eg, via vmap,
+ * vmalloc, ioremap etc) in kernel space for pages.  On non-VIPT
+ * caches, since the direct-mappings of these pages may contain cached
+ * data, we need to do a full cache flush to ensure that writebacks
+ * don't corrupt data placed into these pages via the new mappings.
+ */
+static inline void flush_cache_vmap(unsigned long start, unsigned long end)
+{
+	/*
+	 * set_pte_at() called from vmap_pte_range() does not
+	 * have a DSB after cleaning the cache line.
+	 */
+	dsb();
+}
+
+static inline void flush_cache_vunmap(unsigned long start, unsigned long end)
+{
+}
+
+#endif

arch/arm64/include/asm/cachetype.h

@@ -0,0 +1,48 @@
+/*
+ * Copyright (C) 2012 ARM Ltd.
+ *
+ * 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.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+#ifndef __ASM_CACHETYPE_H
+#define __ASM_CACHETYPE_H
+
+#include <asm/cputype.h>
+
+#define CTR_L1IP_SHIFT		14
+#define CTR_L1IP_MASK		3
+
+#define ICACHE_POLICY_RESERVED	0
+#define ICACHE_POLICY_AIVIVT	1
+#define ICACHE_POLICY_VIPT	2
+#define ICACHE_POLICY_PIPT	3
+
+static inline u32 icache_policy(void)
+{
+	return (read_cpuid_cachetype() >> CTR_L1IP_SHIFT) & CTR_L1IP_MASK;
+}
+
+/*
+ * Whilst the D-side always behaves as PIPT on AArch64, aliasing is
+ * permitted in the I-cache.
+ */
+static inline int icache_is_aliasing(void)
+{
+	return icache_policy() != ICACHE_POLICY_PIPT;
+}
+
+static inline int icache_is_aivivt(void)
+{
+	return icache_policy() == ICACHE_POLICY_AIVIVT;
+}
+
+#endif	/* __ASM_CACHETYPE_H */

arch/arm64/mm/cache.S

@@ -0,0 +1,168 @@
+/*
+ * Cache maintenance
+ *
+ * Copyright (C) 2001 Deep Blue Solutions Ltd.
+ * Copyright (C) 2012 ARM Ltd.
+ *
+ * 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.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/linkage.h>
+#include <linux/init.h>
+#include <asm/assembler.h>
+
+#include "proc-macros.S"
+
+/*
+ *	__flush_dcache_all()
+ *
+ *	Flush the whole D-cache.
+ *
+ *	Corrupted registers: x0-x7, x9-x11
+ */
+ENTRY(__flush_dcache_all)
+	dsb	sy				// ensure ordering with previous memory accesses
+	mrs	x0, clidr_el1			// read clidr
+	and	x3, x0, #0x7000000		// extract loc from clidr
+	lsr	x3, x3, #23			// left align loc bit field
+	cbz	x3, finished			// if loc is 0, then no need to clean
+	mov	x10, #0				// start clean at cache level 0
+loop1:
+	add	x2, x10, x10, lsr #1		// work out 3x current cache level
+	lsr	x1, x0, x2			// extract cache type bits from clidr
+	and	x1, x1, #7			// mask of the bits for current cache only
+	cmp	x1, #2				// see what cache we have at this level
+	b.lt	skip				// skip if no cache, or just i-cache
+	save_and_disable_irqs x9		// make CSSELR and CCSIDR access atomic
+	msr	csselr_el1, x10			// select current cache level in csselr
+	isb					// isb to sych the new cssr&csidr
+	mrs	x1, ccsidr_el1			// read the new ccsidr
+	restore_irqs x9
+	and	x2, x1, #7			// extract the length of the cache lines
+	add	x2, x2, #4			// add 4 (line length offset)
+	mov	x4, #0x3ff
+	and	x4, x4, x1, lsr #3		// find maximum number on the way size
+	clz	x5, x4				// find bit position of way size increment
+	mov	x7, #0x7fff
+	and	x7, x7, x1, lsr #13		// extract max number of the index size
+loop2:
+	mov	x9, x4				// create working copy of max way size
+loop3:
+	lsl	x6, x9, x5
+	orr	x11, x10, x6			// factor way and cache number into x11
+	lsl	x6, x7, x2
+	orr	x11, x11, x6			// factor index number into x11
+	dc	cisw, x11			// clean & invalidate by set/way
+	subs	x9, x9, #1			// decrement the way
+	b.ge	loop3
+	subs	x7, x7, #1			// decrement the index
+	b.ge	loop2
+skip:
+	add	x10, x10, #2			// increment cache number
+	cmp	x3, x10
+	b.gt	loop1
+finished:
+	mov	x10, #0				// swith back to cache level 0
+	msr	csselr_el1, x10			// select current cache level in csselr
+	dsb	sy
+	isb
+	ret
+ENDPROC(__flush_dcache_all)
+
+/*
+ *	flush_cache_all()
+ *
+ *	Flush the entire cache system.  The data cache flush is now achieved
+ *	using atomic clean / invalidates working outwards from L1 cache. This
+ *	is done using Set/Way based cache maintainance instructions.  The
+ *	instruction cache can still be invalidated back to the point of
+ *	unification in a single instruction.
+ */
+ENTRY(flush_cache_all)
+	mov	x12, lr
+	bl	__flush_dcache_all
+	mov	x0, #0
+	ic	ialluis				// I+BTB cache invalidate
+	ret	x12
+ENDPROC(flush_cache_all)
+
+/*
+ *	flush_icache_range(start,end)
+ *
+ *	Ensure that the I and D caches are coherent within specified region.
+ *	This is typically used when code has been written to a memory region,
+ *	and will be executed.
+ *
+ *	- start   - virtual start address of region
+ *	- end     - virtual end address of region
+ */
+ENTRY(flush_icache_range)
+	/* FALLTHROUGH */
+
+/*
+ *	__flush_cache_user_range(start,end)
+ *
+ *	Ensure that the I and D caches are coherent within specified region.
+ *	This is typically used when code has been written to a memory region,
+ *	and will be executed.
+ *
+ *	- start   - virtual start address of region
+ *	- end     - virtual end address of region
+ */
+ENTRY(__flush_cache_user_range)
+	dcache_line_size x2, x3
+	sub	x3, x2, #1
+	bic	x4, x0, x3
+1:
+USER(9f, dc	cvau, x4	)		// clean D line to PoU
+	add	x4, x4, x2
+	cmp	x4, x1
+	b.lo	1b
+	dsb	sy
+
+	icache_line_size x2, x3
+	sub	x3, x2, #1
+	bic	x4, x0, x3
+1:
+USER(9f, ic	ivau, x4	)		// invalidate I line PoU
+	add	x4, x4, x2
+	cmp	x4, x1
+	b.lo	1b
+9:						// ignore any faulting cache operation
+	dsb	sy
+	isb
+	ret
+ENDPROC(flush_icache_range)
+ENDPROC(__flush_cache_user_range)
+
+/*
+ *	__flush_kern_dcache_page(kaddr)
+ *
+ *	Ensure that the data held in the page kaddr is written back to the
+ *	page in question.
+ *
+ *	- kaddr   - kernel address
+ *	- size    - size in question
+ */
+ENTRY(__flush_dcache_area)
+	dcache_line_size x2, x3
+	add	x1, x0, x1
+	sub	x3, x2, #1
+	bic	x0, x0, x3
+1:	dc	civac, x0			// clean & invalidate D line / unified line
+	add	x0, x0, x2
+	cmp	x0, x1
+	b.lo	1b
+	dsb	sy
+	ret
+ENDPROC(__flush_dcache_area)