[linux.git] / include / asm-ia64 / tlb.h

#ifndef _ASM_IA64_TLB_H
#define _ASM_IA64_TLB_H
/*
 * Based on <asm-generic/tlb.h>.
 *
 * Copyright (C) 2002-2003 Hewlett-Packard Co
 *	David Mosberger-Tang <[email protected]>
 */
/*
 * Removing a translation from a page table (including TLB-shootdown) is a four-step
 * procedure:
 *
 *	(1) Flush (virtual) caches --- ensures virtual memory is coherent with kernel memory
 *	    (this is a no-op on ia64).
 *	(2) Clear the relevant portions of the page-table
 *	(3) Flush the TLBs --- ensures that stale content is gone from CPU TLBs
 *	(4) Release the pages that were freed up in step (2).
 *
 * Note that the ordering of these steps is crucial to avoid races on MP machines.
 *
 * The Linux kernel defines several platform-specific hooks for TLB-shootdown.  When
 * unmapping a portion of the virtual address space, these hooks are called according to
 * the following template:
 *
 *	tlb <- tlb_gather_mmu(mm, full_mm_flush);	// start unmap for address space MM
 *	{
 *	  for each vma that needs a shootdown do {
 *	    tlb_start_vma(tlb, vma);
 *	      for each page-table-entry PTE that needs to be removed do {
 *		tlb_remove_tlb_entry(tlb, pte, address);
 *		if (pte refers to a normal page) {
 *		  tlb_remove_page(tlb, page);
 *		}
 *	      }
 *	    tlb_end_vma(tlb, vma);
 *	  }
 *	}
 *	tlb_finish_mmu(tlb, start, end);	// finish unmap for address space MM
 */
#include <linux/config.h>
#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/swap.h>

#include <asm/pgalloc.h>
#include <asm/processor.h>
#include <asm/tlbflush.h>
#include <asm/machvec.h>

#ifdef CONFIG_SMP
# define FREE_PTE_NR		2048
# define tlb_fast_mode(tlb)	((tlb)->nr == ~0U)
#else
# define FREE_PTE_NR		0
# define tlb_fast_mode(tlb)	(1)
#endif

struct mmu_gather {
	struct mm_struct	*mm;
	unsigned int		nr;		/* == ~0U => fast mode */
	unsigned char		fullmm;		/* non-zero means full mm flush */
	unsigned char		need_flush;	/* really unmapped some PTEs? */
	unsigned long		start_addr;
	unsigned long		end_addr;
	struct page 		*pages[FREE_PTE_NR];
};

/* Users of the generic TLB shootdown code must declare this storage space. */
DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);

/*
 * Flush the TLB for address range START to END and, if not in fast mode, release the
 * freed pages that where gathered up to this point.
 */
static inline void
ia64_tlb_flush_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
{
	unsigned int nr;

	if (!tlb->need_flush)
		return;
	tlb->need_flush = 0;

	if (tlb->fullmm) {
		/*
		 * Tearing down the entire address space.  This happens both as a result
		 * of exit() and execve().  The latter case necessitates the call to
		 * flush_tlb_mm() here.
		 */
		flush_tlb_mm(tlb->mm);
	} else if (unlikely (end - start >= 1024*1024*1024*1024UL
			     || REGION_NUMBER(start) != REGION_NUMBER(end - 1)))
	{
		/*
		 * If we flush more than a tera-byte or across regions, we're probably
		 * better off just flushing the entire TLB(s).  This should be very rare
		 * and is not worth optimizing for.
		 */
		flush_tlb_all();
	} else {
		/*
		 * XXX fix me: flush_tlb_range() should take an mm pointer instead of a
		 * vma pointer.
		 */
		struct vm_area_struct vma;

		vma.vm_mm = tlb->mm;
		/* flush the address range from the tlb: */
		flush_tlb_range(&vma, start, end);
		/* now flush the virt. page-table area mapping the address range: */
		flush_tlb_range(&vma, ia64_thash(start), ia64_thash(end));
	}

	/* lastly, release the freed pages */
	nr = tlb->nr;
	if (!tlb_fast_mode(tlb)) {
		unsigned long i;
		tlb->nr = 0;
		tlb->start_addr = ~0UL;
		for (i = 0; i < nr; ++i)
			free_page_and_swap_cache(tlb->pages[i]);
	}
}

/*
 * Return a pointer to an initialized struct mmu_gather.
 */
static inline struct mmu_gather *
tlb_gather_mmu (struct mm_struct *mm, unsigned int full_mm_flush)
{
	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);

	tlb->mm = mm;
	/*
	 * Use fast mode if only 1 CPU is online.
	 *
	 * It would be tempting to turn on fast-mode for full_mm_flush as well.  But this
	 * doesn't work because of speculative accesses and software prefetching: the page
	 * table of "mm" may (and usually is) the currently active page table and even
	 * though the kernel won't do any user-space accesses during the TLB shoot down, a
	 * compiler might use speculation or lfetch.fault on what happens to be a valid
	 * user-space address.  This in turn could trigger a TLB miss fault (or a VHPT
	 * walk) and re-insert a TLB entry we just removed.  Slow mode avoids such
	 * problems.  (We could make fast-mode work by switching the current task to a
	 * different "mm" during the shootdown.) --davidm 08/02/2002
	 */
	tlb->nr = (num_online_cpus() == 1) ? ~0U : 0;
	tlb->fullmm = full_mm_flush;
	tlb->start_addr = ~0UL;
	return tlb;
}

/*
 * Called at the end of the shootdown operation to free up any resources that were
 * collected.
 */
static inline void
tlb_finish_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
{
	/*
	 * Note: tlb->nr may be 0 at this point, so we can't rely on tlb->start_addr and
	 * tlb->end_addr.
	 */
	ia64_tlb_flush_mmu(tlb, start, end);

	/* keep the page table cache within bounds */
	check_pgt_cache();

	put_cpu_var(mmu_gathers);
}

/*
 * Logically, this routine frees PAGE.  On MP machines, the actual freeing of the page
 * must be delayed until after the TLB has been flushed (see comments at the beginning of
 * this file).
 */
static inline void
tlb_remove_page (struct mmu_gather *tlb, struct page *page)
{
	tlb->need_flush = 1;

	if (tlb_fast_mode(tlb)) {
		free_page_and_swap_cache(page);
		return;
	}
	tlb->pages[tlb->nr++] = page;
	if (tlb->nr >= FREE_PTE_NR)
		ia64_tlb_flush_mmu(tlb, tlb->start_addr, tlb->end_addr);
}

/*
 * Remove TLB entry for PTE mapped at virtual address ADDRESS.  This is called for any
 * PTE, not just those pointing to (normal) physical memory.
 */
static inline void
__tlb_remove_tlb_entry (struct mmu_gather *tlb, pte_t *ptep, unsigned long address)
{
	if (tlb->start_addr == ~0UL)
		tlb->start_addr = address;
	tlb->end_addr = address + PAGE_SIZE;
}

#define tlb_migrate_finish(mm)	platform_tlb_migrate_finish(mm)

#define tlb_start_vma(tlb, vma)			do { } while (0)
#define tlb_end_vma(tlb, vma)			do { } while (0)

#define tlb_remove_tlb_entry(tlb, ptep, addr)		\
do {							\
	tlb->need_flush = 1;				\
	__tlb_remove_tlb_entry(tlb, ptep, addr);	\
} while (0)

#define pte_free_tlb(tlb, ptep)				\
do {							\
	tlb->need_flush = 1;				\
	__pte_free_tlb(tlb, ptep);			\
} while (0)

#define pmd_free_tlb(tlb, ptep)				\
do {							\
	tlb->need_flush = 1;				\
	__pmd_free_tlb(tlb, ptep);			\
} while (0)

#define pud_free_tlb(tlb, pudp)				\
do {							\
	tlb->need_flush = 1;				\
	__pud_free_tlb(tlb, pudp);			\
} while (0)

#endif /* _ASM_IA64_TLB_H */
Commit	Line	Data
1da177e4 LT	1	#ifndef _ASM_IA64_TLB_H
	2	#define _ASM_IA64_TLB_H
	3	/*
	4	* Based on <asm-generic/tlb.h>.
	5	*
	6	* Copyright (C) 2002-2003 Hewlett-Packard Co
	7	* David Mosberger-Tang <[email protected]>
	8	*/
	9	/*
	10	* Removing a translation from a page table (including TLB-shootdown) is a four-step
	11	* procedure:
	12	*
	13	* (1) Flush (virtual) caches --- ensures virtual memory is coherent with kernel memory
	14	* (this is a no-op on ia64).
	15	* (2) Clear the relevant portions of the page-table
	16	* (3) Flush the TLBs --- ensures that stale content is gone from CPU TLBs
	17	* (4) Release the pages that were freed up in step (2).
	18	*
	19	* Note that the ordering of these steps is crucial to avoid races on MP machines.
	20	*
	21	* The Linux kernel defines several platform-specific hooks for TLB-shootdown. When
	22	* unmapping a portion of the virtual address space, these hooks are called according to
	23	* the following template:
	24	*
	25	* tlb <- tlb_gather_mmu(mm, full_mm_flush); // start unmap for address space MM
	26	* {
	27	* for each vma that needs a shootdown do {
	28	* tlb_start_vma(tlb, vma);
	29	* for each page-table-entry PTE that needs to be removed do {
	30	* tlb_remove_tlb_entry(tlb, pte, address);
	31	* if (pte refers to a normal page) {
	32	* tlb_remove_page(tlb, page);
	33	* }
	34	* }
	35	* tlb_end_vma(tlb, vma);
	36	* }
	37	* }
	38	* tlb_finish_mmu(tlb, start, end); // finish unmap for address space MM
	39	*/
	40	#include <linux/config.h>
	41	#include <linux/mm.h>
	42	#include <linux/pagemap.h>
	43	#include <linux/swap.h>
	44
	45	#include <asm/pgalloc.h>
	46	#include <asm/processor.h>
	47	#include <asm/tlbflush.h>
	48	#include <asm/machvec.h>
	49
	50	#ifdef CONFIG_SMP
	51	# define FREE_PTE_NR 2048
	52	# define tlb_fast_mode(tlb) ((tlb)->nr == ~0U)
	53	#else
	54	# define FREE_PTE_NR 0
	55	# define tlb_fast_mode(tlb) (1)
	56	#endif
	57
	58	struct mmu_gather {
	59	struct mm_struct *mm;
	60	unsigned int nr; /* == ~0U => fast mode */
	61	unsigned char fullmm; /* non-zero means full mm flush */
	62	unsigned char need_flush; /* really unmapped some PTEs? */
1da177e4 LT	63	unsigned long start_addr;
	64	unsigned long end_addr;
	65	struct page *pages[FREE_PTE_NR];
	66	};
	67
	68	/* Users of the generic TLB shootdown code must declare this storage space. */
	69	DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
	70
	71	/*
	72	* Flush the TLB for address range START to END and, if not in fast mode, release the
	73	* freed pages that where gathered up to this point.
	74	*/
	75	static inline void
	76	ia64_tlb_flush_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
	77	{
	78	unsigned int nr;
	79
	80	if (!tlb->need_flush)
	81	return;
	82	tlb->need_flush = 0;
	83
	84	if (tlb->fullmm) {
	85	/*
	86	* Tearing down the entire address space. This happens both as a result
	87	* of exit() and execve(). The latter case necessitates the call to
	88	* flush_tlb_mm() here.
	89	*/
	90	flush_tlb_mm(tlb->mm);
	91	} else if (unlikely (end - start >= 102410241024*1024UL
	92	\|\| REGION_NUMBER(start) != REGION_NUMBER(end - 1)))
	93	{
	94	/*
	95	* If we flush more than a tera-byte or across regions, we're probably
	96	* better off just flushing the entire TLB(s). This should be very rare
	97	* and is not worth optimizing for.
	98	*/
	99	flush_tlb_all();
	100	} else {
	101	/*
	102	* XXX fix me: flush_tlb_range() should take an mm pointer instead of a
	103	* vma pointer.
	104	*/
	105	struct vm_area_struct vma;
	106
	107	vma.vm_mm = tlb->mm;
	108	/* flush the address range from the tlb: */
	109	flush_tlb_range(&vma, start, end);
	110	/* now flush the virt. page-table area mapping the address range: */
	111	flush_tlb_range(&vma, ia64_thash(start), ia64_thash(end));
	112	}
	113
	114	/* lastly, release the freed pages */
	115	nr = tlb->nr;
	116	if (!tlb_fast_mode(tlb)) {
	117	unsigned long i;
	118	tlb->nr = 0;
	119	tlb->start_addr = ~0UL;
	120	for (i = 0; i < nr; ++i)
	121	free_page_and_swap_cache(tlb->pages[i]);
	122	}
	123	}
	124
	125	/*
	126	* Return a pointer to an initialized struct mmu_gather.
127	*/
128	static inline struct mmu_gather *
129	tlb_gather_mmu (struct mm_struct *mm, unsigned int full_mm_flush)
130	{
15a23ffa	131	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
1da177e4 LT	132
	133	tlb->mm = mm;
	134	/*
	135	* Use fast mode if only 1 CPU is online.
	136	*
	137	* It would be tempting to turn on fast-mode for full_mm_flush as well. But this
	138	* doesn't work because of speculative accesses and software prefetching: the page
	139	* table of "mm" may (and usually is) the currently active page table and even
	140	* though the kernel won't do any user-space accesses during the TLB shoot down, a
	141	* compiler might use speculation or lfetch.fault on what happens to be a valid
	142	* user-space address. This in turn could trigger a TLB miss fault (or a VHPT
	143	* walk) and re-insert a TLB entry we just removed. Slow mode avoids such
	144	* problems. (We could make fast-mode work by switching the current task to a
	145	* different "mm" during the shootdown.) --davidm 08/02/2002
	146	*/
	147	tlb->nr = (num_online_cpus() == 1) ? ~0U : 0;
	148	tlb->fullmm = full_mm_flush;
1da177e4 LT	149	tlb->start_addr = ~0UL;
	150	return tlb;
	151	}
	152
	153	/*
	154	* Called at the end of the shootdown operation to free up any resources that were
15a23ffa	155	* collected.
1da177e4 LT	156	*/
	157	static inline void
	158	tlb_finish_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
	159	{
1da177e4 LT	160	/*
	161	* Note: tlb->nr may be 0 at this point, so we can't rely on tlb->start_addr and
	162	* tlb->end_addr.
	163	*/
	164	ia64_tlb_flush_mmu(tlb, start, end);
	165
	166	/* keep the page table cache within bounds */
	167	check_pgt_cache();
15a23ffa HD	168
15a23ffa HD	169	put_cpu_var(mmu_gathers);
1da177e4 LT	170	}
1da177e4 LT	171
1da177e4 LT	172	/*
	173	* Logically, this routine frees PAGE. On MP machines, the actual freeing of the page
	174	* must be delayed until after the TLB has been flushed (see comments at the beginning of
	175	* this file).
	176	*/
	177	static inline void
	178	tlb_remove_page (struct mmu_gather tlb, struct page page)
	179	{
	180	tlb->need_flush = 1;
	181
	182	if (tlb_fast_mode(tlb)) {
	183	free_page_and_swap_cache(page);
	184	return;
	185	}
	186	tlb->pages[tlb->nr++] = page;
	187	if (tlb->nr >= FREE_PTE_NR)
	188	ia64_tlb_flush_mmu(tlb, tlb->start_addr, tlb->end_addr);
	189	}
	190
	191	/*
	192	* Remove TLB entry for PTE mapped at virtual address ADDRESS. This is called for any
	193	* PTE, not just those pointing to (normal) physical memory.
	194	*/
	195	static inline void
	196	__tlb_remove_tlb_entry (struct mmu_gather tlb, pte_t ptep, unsigned long address)
	197	{
	198	if (tlb->start_addr == ~0UL)
	199	tlb->start_addr = address;
	200	tlb->end_addr = address + PAGE_SIZE;
	201	}
	202
	203	#define tlb_migrate_finish(mm) platform_tlb_migrate_finish(mm)
	204
	205	#define tlb_start_vma(tlb, vma) do { } while (0)
	206	#define tlb_end_vma(tlb, vma) do { } while (0)
	207
	208	#define tlb_remove_tlb_entry(tlb, ptep, addr) \
	209	do { \
	210	tlb->need_flush = 1; \
	211	__tlb_remove_tlb_entry(tlb, ptep, addr); \
	212	} while (0)
	213
	214	#define pte_free_tlb(tlb, ptep) \
	215	do { \
	216	tlb->need_flush = 1; \
	217	__pte_free_tlb(tlb, ptep); \
	218	} while (0)
	219
	220	#define pmd_free_tlb(tlb, ptep) \
	221	do { \
	222	tlb->need_flush = 1; \
	223	__pmd_free_tlb(tlb, ptep); \
	224	} while (0)
	225
	226	#define pud_free_tlb(tlb, pudp) \
	227	do { \
	228	tlb->need_flush = 1; \
	229	__pud_free_tlb(tlb, pudp); \
	230	} while (0)
	231
	232	#endif /* _ASM_IA64_TLB_H */