From 9a8fd5589902153a134111ed7a40f9cca1f83254 Mon Sep 17 00:00:00 2001 From: Chris Zankel Date: Thu, 23 Jun 2005 22:01:26 -0700 Subject: [PATCH] xtensa: Architecture support for Tensilica Xtensa Part 6 The attached patches provides part 6 of an architecture implementation for the Tensilica Xtensa CPU series. Signed-off-by: Chris Zankel Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- include/asm-xtensa/mmu_context.h | 330 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 330 insertions(+) create mode 100644 include/asm-xtensa/mmu_context.h (limited to 'include/asm-xtensa/mmu_context.h') diff --git a/include/asm-xtensa/mmu_context.h b/include/asm-xtensa/mmu_context.h new file mode 100644 index 0000000..1b08015 --- /dev/null +++ b/include/asm-xtensa/mmu_context.h @@ -0,0 +1,330 @@ +/* + * include/asm-xtensa/mmu_context.h + * + * Switch an MMU context. + * + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + * + * Copyright (C) 2001 - 2005 Tensilica Inc. + */ + +#ifndef _XTENSA_MMU_CONTEXT_H +#define _XTENSA_MMU_CONTEXT_H + +#include +#include + +#include +#include +#include +#include + +/* + * Linux was ported to Xtensa assuming all auto-refill ways in set 0 + * had the same properties (a very likely assumption). Multiple sets + * of auto-refill ways will still work properly, but not as optimally + * as the Xtensa designer may have assumed. + * + * We make this case a hard #error, killing the kernel build, to alert + * the developer to this condition (which is more likely an error). + * You super-duper clever developers can change it to a warning or + * remove it altogether if you think you know what you're doing. :) + */ + +#if (XCHAL_HAVE_TLBS != 1) +# error "Linux must have an MMU!" +#endif + +#if ((XCHAL_ITLB_ARF_WAYS == 0) || (XCHAL_DTLB_ARF_WAYS == 0)) +# error "MMU must have auto-refill ways" +#endif + +#if ((XCHAL_ITLB_ARF_SETS != 1) || (XCHAL_DTLB_ARF_SETS != 1)) +# error Linux may not use all auto-refill ways as efficiently as you think +#endif + +#if (XCHAL_MMU_MAX_PTE_PAGE_SIZE != XCHAL_MMU_MIN_PTE_PAGE_SIZE) +# error Only one page size allowed! +#endif + +extern unsigned long asid_cache; +extern pgd_t *current_pgd; + +/* + * Define the number of entries per auto-refill way in set 0 of both I and D + * TLBs. We deal only with set 0 here (an assumption further explained in + * assertions.h). Also, define the total number of ARF entries in both TLBs. + */ + +#define ITLB_ENTRIES_PER_ARF_WAY (XCHAL_ITLB_SET(XCHAL_ITLB_ARF_SET0,ENTRIES)) +#define DTLB_ENTRIES_PER_ARF_WAY (XCHAL_DTLB_SET(XCHAL_DTLB_ARF_SET0,ENTRIES)) + +#define ITLB_ENTRIES \ + (ITLB_ENTRIES_PER_ARF_WAY * (XCHAL_ITLB_SET(XCHAL_ITLB_ARF_SET0,WAYS))) +#define DTLB_ENTRIES \ + (DTLB_ENTRIES_PER_ARF_WAY * (XCHAL_DTLB_SET(XCHAL_DTLB_ARF_SET0,WAYS))) + + +/* + * SMALLEST_NTLB_ENTRIES is the smaller of ITLB_ENTRIES and DTLB_ENTRIES. + * In practice, they are probably equal. This macro simplifies function + * flush_tlb_range(). + */ + +#if (DTLB_ENTRIES < ITLB_ENTRIES) +# define SMALLEST_NTLB_ENTRIES DTLB_ENTRIES +#else +# define SMALLEST_NTLB_ENTRIES ITLB_ENTRIES +#endif + + +/* + * asid_cache tracks only the ASID[USER_RING] field of the RASID special + * register, which is the current user-task asid allocation value. + * mm->context has the same meaning. When it comes time to write the + * asid_cache or mm->context values to the RASID special register, we first + * shift the value left by 8, then insert the value. + * ASID[0] always contains the kernel's asid value, and we reserve three + * other asid values that we never assign to user tasks. + */ + +#define ASID_INC 0x1 +#define ASID_MASK ((1 << XCHAL_MMU_ASID_BITS) - 1) + +/* + * XCHAL_MMU_ASID_INVALID is a configurable Xtensa processor constant + * indicating invalid address space. XCHAL_MMU_ASID_KERNEL is a configurable + * Xtensa processor constant indicating the kernel address space. They can + * be arbitrary values. + * + * We identify three more unique, reserved ASID values to use in the unused + * ring positions. No other user process will be assigned these reserved + * ASID values. + * + * For example, given that + * + * XCHAL_MMU_ASID_INVALID == 0 + * XCHAL_MMU_ASID_KERNEL == 1 + * + * the following maze of #if statements would generate + * + * ASID_RESERVED_1 == 2 + * ASID_RESERVED_2 == 3 + * ASID_RESERVED_3 == 4 + * ASID_FIRST_NONRESERVED == 5 + */ + +#if (XCHAL_MMU_ASID_INVALID != XCHAL_MMU_ASID_KERNEL + 1) +# define ASID_RESERVED_1 ((XCHAL_MMU_ASID_KERNEL + 1) & ASID_MASK) +#else +# define ASID_RESERVED_1 ((XCHAL_MMU_ASID_KERNEL + 2) & ASID_MASK) +#endif + +#if (XCHAL_MMU_ASID_INVALID != ASID_RESERVED_1 + 1) +# define ASID_RESERVED_2 ((ASID_RESERVED_1 + 1) & ASID_MASK) +#else +# define ASID_RESERVED_2 ((ASID_RESERVED_1 + 2) & ASID_MASK) +#endif + +#if (XCHAL_MMU_ASID_INVALID != ASID_RESERVED_2 + 1) +# define ASID_RESERVED_3 ((ASID_RESERVED_2 + 1) & ASID_MASK) +#else +# define ASID_RESERVED_3 ((ASID_RESERVED_2 + 2) & ASID_MASK) +#endif + +#if (XCHAL_MMU_ASID_INVALID != ASID_RESERVED_3 + 1) +# define ASID_FIRST_NONRESERVED ((ASID_RESERVED_3 + 1) & ASID_MASK) +#else +# define ASID_FIRST_NONRESERVED ((ASID_RESERVED_3 + 2) & ASID_MASK) +#endif + +#define ASID_ALL_RESERVED ( ((ASID_RESERVED_1) << 24) + \ + ((ASID_RESERVED_2) << 16) + \ + ((ASID_RESERVED_3) << 8) + \ + ((XCHAL_MMU_ASID_KERNEL)) ) + + +/* + * NO_CONTEXT is the invalid ASID value that we don't ever assign to + * any user or kernel context. NO_CONTEXT is a better mnemonic than + * XCHAL_MMU_ASID_INVALID, so we use it in code instead. + */ + +#define NO_CONTEXT XCHAL_MMU_ASID_INVALID + +#if (KERNEL_RING != 0) +# error The KERNEL_RING really should be zero. +#endif + +#if (USER_RING >= XCHAL_MMU_RINGS) +# error USER_RING cannot be greater than the highest numbered ring. +#endif + +#if (USER_RING == KERNEL_RING) +# error The user and kernel rings really should not be equal. +#endif + +#if (USER_RING == 1) +#define ASID_INSERT(x) ( ((ASID_RESERVED_1) << 24) + \ + ((ASID_RESERVED_2) << 16) + \ + (((x) & (ASID_MASK)) << 8) + \ + ((XCHAL_MMU_ASID_KERNEL)) ) + +#elif (USER_RING == 2) +#define ASID_INSERT(x) ( ((ASID_RESERVED_1) << 24) + \ + (((x) & (ASID_MASK)) << 16) + \ + ((ASID_RESERVED_2) << 8) + \ + ((XCHAL_MMU_ASID_KERNEL)) ) + +#elif (USER_RING == 3) +#define ASID_INSERT(x) ( (((x) & (ASID_MASK)) << 24) + \ + ((ASID_RESERVED_1) << 16) + \ + ((ASID_RESERVED_2) << 8) + \ + ((XCHAL_MMU_ASID_KERNEL)) ) + +#else +#error Goofy value for USER_RING + +#endif /* USER_RING == 1 */ + + +/* + * All unused by hardware upper bits will be considered + * as a software asid extension. + */ + +#define ASID_VERSION_MASK ((unsigned long)~(ASID_MASK|(ASID_MASK-1))) +#define ASID_FIRST_VERSION \ + ((unsigned long)(~ASID_VERSION_MASK) + 1 + ASID_FIRST_NONRESERVED) + +extern inline void set_rasid_register (unsigned long val) +{ + __asm__ __volatile__ (" wsr %0, "__stringify(RASID)"\n\t" + " isync\n" : : "a" (val)); +} + +extern inline unsigned long get_rasid_register (void) +{ + unsigned long tmp; + __asm__ __volatile__ (" rsr %0, "__stringify(RASID)"\n\t" : "=a" (tmp)); + return tmp; +} + + +#if ((XCHAL_MMU_ASID_INVALID == 0) && (XCHAL_MMU_ASID_KERNEL == 1)) + +extern inline void +get_new_mmu_context(struct mm_struct *mm, unsigned long asid) +{ + extern void flush_tlb_all(void); + if (! ((asid += ASID_INC) & ASID_MASK) ) { + flush_tlb_all(); /* start new asid cycle */ + if (!asid) /* fix version if needed */ + asid = ASID_FIRST_VERSION - ASID_FIRST_NONRESERVED; + asid += ASID_FIRST_NONRESERVED; + } + mm->context = asid_cache = asid; +} + +#else +#warning ASID_{INVALID,KERNEL} values impose non-optimal get_new_mmu_context implementation + +/* XCHAL_MMU_ASID_INVALID == 0 and XCHAL_MMU_ASID_KERNEL ==1 are + really the best, but if you insist... */ + +extern inline int validate_asid (unsigned long asid) +{ + switch (asid) { + case XCHAL_MMU_ASID_INVALID: + case XCHAL_MMU_ASID_KERNEL: + case ASID_RESERVED_1: + case ASID_RESERVED_2: + case ASID_RESERVED_3: + return 0; /* can't use these values as ASIDs */ + } + return 1; /* valid */ +} + +extern inline void +get_new_mmu_context(struct mm_struct *mm, unsigned long asid) +{ + extern void flush_tlb_all(void); + while (1) { + asid += ASID_INC; + if ( ! (asid & ASID_MASK) ) { + flush_tlb_all(); /* start new asid cycle */ + if (!asid) /* fix version if needed */ + asid = ASID_FIRST_VERSION - ASID_FIRST_NONRESERVED; + asid += ASID_FIRST_NONRESERVED; + break; /* no need to validate here */ + } + if (validate_asid (asid & ASID_MASK)) + break; + } + mm->context = asid_cache = asid; +} + +#endif + + +/* + * Initialize the context related info for a new mm_struct + * instance. + */ + +extern inline int +init_new_context(struct task_struct *tsk, struct mm_struct *mm) +{ + mm->context = NO_CONTEXT; + return 0; +} + +extern inline void switch_mm(struct mm_struct *prev, struct mm_struct *next, + struct task_struct *tsk) +{ + unsigned long asid = asid_cache; + + /* Check if our ASID is of an older version and thus invalid */ + + if ((next->context ^ asid) & ASID_VERSION_MASK) + get_new_mmu_context(next, asid); + + set_rasid_register (ASID_INSERT(next->context)); + invalidate_page_directory(); +} + +#define deactivate_mm(tsk, mm) do { } while(0) + +/* + * Destroy context related info for an mm_struct that is about + * to be put to rest. + */ +extern inline void destroy_context(struct mm_struct *mm) +{ + /* Nothing to do. */ +} + +/* + * After we have set current->mm to a new value, this activates + * the context for the new mm so we see the new mappings. + */ +extern inline void +activate_mm(struct mm_struct *prev, struct mm_struct *next) +{ + /* Unconditionally get a new ASID. */ + + get_new_mmu_context(next, asid_cache); + set_rasid_register (ASID_INSERT(next->context)); + invalidate_page_directory(); +} + + +static inline void enter_lazy_tlb(struct mm_struct *mm, struct task_struct *tsk) +{ + /* Nothing to do. */ + +} + +#endif /* _XTENSA_MMU_CONTEXT_H */ -- cgit v1.1