17c478bd9Sstevel@tonic-gate /* 27c478bd9Sstevel@tonic-gate * CDDL HEADER START 37c478bd9Sstevel@tonic-gate * 47c478bd9Sstevel@tonic-gate * The contents of this file are subject to the terms of the 5a985e578Skchow * Common Development and Distribution License (the "License"). 6a985e578Skchow * You may not use this file except in compliance with the License. 77c478bd9Sstevel@tonic-gate * 87c478bd9Sstevel@tonic-gate * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 97c478bd9Sstevel@tonic-gate * or http://www.opensolaris.org/os/licensing. 107c478bd9Sstevel@tonic-gate * See the License for the specific language governing permissions 117c478bd9Sstevel@tonic-gate * and limitations under the License. 127c478bd9Sstevel@tonic-gate * 137c478bd9Sstevel@tonic-gate * When distributing Covered Code, include this CDDL HEADER in each 147c478bd9Sstevel@tonic-gate * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 157c478bd9Sstevel@tonic-gate * If applicable, add the following below this CDDL HEADER, with the 167c478bd9Sstevel@tonic-gate * fields enclosed by brackets "[]" replaced with your own identifying 177c478bd9Sstevel@tonic-gate * information: Portions Copyright [yyyy] [name of copyright owner] 187c478bd9Sstevel@tonic-gate * 197c478bd9Sstevel@tonic-gate * CDDL HEADER END 207c478bd9Sstevel@tonic-gate */ 217c478bd9Sstevel@tonic-gate /* 22*b52a336eSPavel Tatashin * Copyright (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved. 237c478bd9Sstevel@tonic-gate */ 24a3114836SGerry Liu /* 25a3114836SGerry Liu * Copyright (c) 2010, Intel Corporation. 26a3114836SGerry Liu * All rights reserved. 27a3114836SGerry Liu */ 287c478bd9Sstevel@tonic-gate 297c478bd9Sstevel@tonic-gate /* 307c478bd9Sstevel@tonic-gate * UNIX machine dependent virtual memory support. 317c478bd9Sstevel@tonic-gate */ 327c478bd9Sstevel@tonic-gate 337c478bd9Sstevel@tonic-gate #ifndef _VM_DEP_H 347c478bd9Sstevel@tonic-gate #define _VM_DEP_H 357c478bd9Sstevel@tonic-gate 367c478bd9Sstevel@tonic-gate 377c478bd9Sstevel@tonic-gate #ifdef __cplusplus 387c478bd9Sstevel@tonic-gate extern "C" { 397c478bd9Sstevel@tonic-gate #endif 407c478bd9Sstevel@tonic-gate 417c478bd9Sstevel@tonic-gate #include <sys/clock.h> 427c478bd9Sstevel@tonic-gate #include <vm/hat_pte.h> 43ae115bc7Smrj #include <sys/param.h> 44ce8eb11aSdp78419 #include <sys/memnode.h> 457c478bd9Sstevel@tonic-gate 467c478bd9Sstevel@tonic-gate /* 47843e1988Sjohnlev * WARNING: vm_dep.h is included by files in common. 487c478bd9Sstevel@tonic-gate */ 497c478bd9Sstevel@tonic-gate 507c478bd9Sstevel@tonic-gate #define GETTICK() tsc_read() 5106fb6a36Sdv142724 /* 5206fb6a36Sdv142724 * Do not use this function for obtaining clock tick. This 5306fb6a36Sdv142724 * is called by callers who do not need to have a guarenteed 5406fb6a36Sdv142724 * correct tick value. The proper routine to use is tsc_read(). 5506fb6a36Sdv142724 */ 567c478bd9Sstevel@tonic-gate 57*b52a336eSPavel Tatashin extern u_longlong_t randtick(); 58843e1988Sjohnlev extern uint_t page_create_update_flags_x86(uint_t); 597c478bd9Sstevel@tonic-gate 60843e1988Sjohnlev extern size_t plcnt_sz(size_t); 61843e1988Sjohnlev #define PLCNT_SZ(ctrs_sz) (ctrs_sz = plcnt_sz(ctrs_sz)) 627c478bd9Sstevel@tonic-gate 63843e1988Sjohnlev extern caddr_t plcnt_init(caddr_t); 64843e1988Sjohnlev #define PLCNT_INIT(addr) (addr = plcnt_init(addr)) 65843e1988Sjohnlev 66843e1988Sjohnlev extern void plcnt_inc_dec(page_t *, int, int, long, int); 67843e1988Sjohnlev #define PLCNT_INCR(pp, mnode, mtype, szc, flags) \ 68843e1988Sjohnlev plcnt_inc_dec(pp, mtype, szc, 1l << PAGE_BSZS_SHIFT(szc), flags) 69843e1988Sjohnlev #define PLCNT_DECR(pp, mnode, mtype, szc, flags) \ 70843e1988Sjohnlev plcnt_inc_dec(pp, mtype, szc, -1l << PAGE_BSZS_SHIFT(szc), flags) 7107ad560dSkchow 727c478bd9Sstevel@tonic-gate /* 73843e1988Sjohnlev * macro to update page list max counts. no-op on x86. 74affbd3ccSkchow */ 75e21bae1bSkchow #define PLCNT_XFER_NORELOC(pp) 76e21bae1bSkchow 77a3114836SGerry Liu /* 78a3114836SGerry Liu * macro to modify the page list max counts when memory is added to 79a3114836SGerry Liu * the page lists during startup (add_physmem) or during a DR operation 80a3114836SGerry Liu * when memory is added (kphysm_add_memory_dynamic) or deleted 81a3114836SGerry Liu * (kphysm_del_cleanup). 82a3114836SGerry Liu */ 83a3114836SGerry Liu #define PLCNT_MODIFY_MAX(pfn, cnt) mtype_modify_max(pfn, cnt) 84a3114836SGerry Liu 857c478bd9Sstevel@tonic-gate extern int memrange_num(pfn_t); 867c478bd9Sstevel@tonic-gate extern int pfn_2_mtype(pfn_t); 877c478bd9Sstevel@tonic-gate extern int mtype_func(int, int, uint_t); 88e21bae1bSkchow extern void mtype_modify_max(pfn_t, long); 89affbd3ccSkchow extern int mnode_pgcnt(int); 905d07b933Sdp78419 extern int mnode_range_cnt(int); 917c478bd9Sstevel@tonic-gate 927c478bd9Sstevel@tonic-gate /* 935d07b933Sdp78419 * candidate counters in vm_pagelist.c are indexed by color and range 945d07b933Sdp78419 */ 95843e1988Sjohnlev #define NUM_MEM_RANGES 4 /* memory range types */ 965d07b933Sdp78419 #define MAX_MNODE_MRANGES NUM_MEM_RANGES 975d07b933Sdp78419 #define MNODE_RANGE_CNT(mnode) mnode_range_cnt(mnode) 98843e1988Sjohnlev #define MNODE_MAX_MRANGE(mnode) memrange_num(mem_node_config[mnode].physbase) 99843e1988Sjohnlev 100843e1988Sjohnlev /* 101843e1988Sjohnlev * This was really badly defined, it implicitly uses mnode_maxmrange[] 102843e1988Sjohnlev * which is a static in vm_pagelist.c 103843e1988Sjohnlev */ 104843e1988Sjohnlev extern int mtype_2_mrange(int); 1055d07b933Sdp78419 #define MTYPE_2_MRANGE(mnode, mtype) \ 106843e1988Sjohnlev (mnode_maxmrange[mnode] - mtype_2_mrange(mtype)) 1075d07b933Sdp78419 1085d07b933Sdp78419 /* 109d94ffb28Sjmcp * Per page size free lists. Allocated dynamically. 1107c478bd9Sstevel@tonic-gate * dimensions [mtype][mmu_page_sizes][colors] 1117c478bd9Sstevel@tonic-gate * 1127c478bd9Sstevel@tonic-gate * mtype specifies a physical memory range with a unique mnode. 1137c478bd9Sstevel@tonic-gate */ 1147c478bd9Sstevel@tonic-gate 115d94ffb28Sjmcp extern page_t ****page_freelists; 1167c478bd9Sstevel@tonic-gate 117d94ffb28Sjmcp #define PAGE_FREELISTS(mnode, szc, color, mtype) \ 118d94ffb28Sjmcp (*(page_freelists[mtype][szc] + (color))) 1197c478bd9Sstevel@tonic-gate 1207c478bd9Sstevel@tonic-gate /* 1217c478bd9Sstevel@tonic-gate * For now there is only a single size cache list. Allocated dynamically. 1227c478bd9Sstevel@tonic-gate * dimensions [mtype][colors] 1237c478bd9Sstevel@tonic-gate * 1247c478bd9Sstevel@tonic-gate * mtype specifies a physical memory range with a unique mnode. 1257c478bd9Sstevel@tonic-gate */ 1267c478bd9Sstevel@tonic-gate extern page_t ***page_cachelists; 1277c478bd9Sstevel@tonic-gate 1287c478bd9Sstevel@tonic-gate #define PAGE_CACHELISTS(mnode, color, mtype) \ 1297c478bd9Sstevel@tonic-gate (*(page_cachelists[mtype] + (color))) 1307c478bd9Sstevel@tonic-gate 1317c478bd9Sstevel@tonic-gate /* 132d94ffb28Sjmcp * There are mutexes for both the page freelist 1337c478bd9Sstevel@tonic-gate * and the page cachelist. We want enough locks to make contention 1347c478bd9Sstevel@tonic-gate * reasonable, but not too many -- otherwise page_freelist_lock() gets 1357c478bd9Sstevel@tonic-gate * so expensive that it becomes the bottleneck! 1367c478bd9Sstevel@tonic-gate */ 1377c478bd9Sstevel@tonic-gate 1387c478bd9Sstevel@tonic-gate #define NPC_MUTEX 16 1397c478bd9Sstevel@tonic-gate 1407c478bd9Sstevel@tonic-gate extern kmutex_t *fpc_mutex[NPC_MUTEX]; 1417c478bd9Sstevel@tonic-gate extern kmutex_t *cpc_mutex[NPC_MUTEX]; 1427c478bd9Sstevel@tonic-gate 143d94ffb28Sjmcp extern page_t *page_get_mnode_freelist(int, uint_t, int, uchar_t, uint_t); 1447c478bd9Sstevel@tonic-gate extern page_t *page_get_mnode_cachelist(uint_t, uint_t, int, int); 1457c478bd9Sstevel@tonic-gate 146ce8eb11aSdp78419 /* mem node iterator is not used on x86 */ 147ce8eb11aSdp78419 #define MEM_NODE_ITERATOR_DECL(it) 148b779d3e0Sdp78419 #define MEM_NODE_ITERATOR_INIT(pfn, mnode, szc, it) 149ce8eb11aSdp78419 150ce8eb11aSdp78419 /* 151ce8eb11aSdp78419 * interleaved_mnodes mode is never set on x86, therefore, 152ce8eb11aSdp78419 * simply return the limits of the given mnode, which then 153ce8eb11aSdp78419 * determines the length of hpm_counters array for the mnode. 154ce8eb11aSdp78419 */ 155ce8eb11aSdp78419 #define HPM_COUNTERS_LIMITS(mnode, physbase, physmax, first) \ 156ce8eb11aSdp78419 { \ 157ce8eb11aSdp78419 (physbase) = mem_node_config[(mnode)].physbase; \ 158ce8eb11aSdp78419 (physmax) = mem_node_config[(mnode)].physmax; \ 159ce8eb11aSdp78419 (first) = (mnode); \ 160ce8eb11aSdp78419 } 161ce8eb11aSdp78419 162ce8eb11aSdp78419 #define PAGE_CTRS_WRITE_LOCK(mnode) \ 163ce8eb11aSdp78419 { \ 164ce8eb11aSdp78419 rw_enter(&page_ctrs_rwlock[(mnode)], RW_WRITER);\ 165ce8eb11aSdp78419 page_freelist_lock(mnode); \ 166ce8eb11aSdp78419 } 167ce8eb11aSdp78419 168ce8eb11aSdp78419 #define PAGE_CTRS_WRITE_UNLOCK(mnode) \ 169ce8eb11aSdp78419 { \ 170ce8eb11aSdp78419 page_freelist_unlock(mnode); \ 171ce8eb11aSdp78419 rw_exit(&page_ctrs_rwlock[(mnode)]); \ 172ce8eb11aSdp78419 } 173ce8eb11aSdp78419 174a3114836SGerry Liu /* 175a3114836SGerry Liu * macro to call page_ctrs_adjust() when memory is added 176a3114836SGerry Liu * during a DR operation. 177a3114836SGerry Liu */ 178a3114836SGerry Liu #define PAGE_CTRS_ADJUST(pfn, cnt, rv) { \ 179a3114836SGerry Liu spgcnt_t _cnt = (spgcnt_t)(cnt); \ 180a3114836SGerry Liu int _mn; \ 181a3114836SGerry Liu pgcnt_t _np; \ 182a3114836SGerry Liu pfn_t _pfn = (pfn); \ 183a3114836SGerry Liu pfn_t _endpfn = _pfn + _cnt; \ 184a3114836SGerry Liu while (_pfn < _endpfn) { \ 185a3114836SGerry Liu _mn = PFN_2_MEM_NODE(_pfn); \ 186a3114836SGerry Liu _np = MIN(_endpfn, mem_node_config[_mn].physmax + 1) - _pfn; \ 187a3114836SGerry Liu _pfn += _np; \ 188a3114836SGerry Liu if ((rv = page_ctrs_adjust(_mn)) != 0) \ 189a3114836SGerry Liu break; \ 190a3114836SGerry Liu } \ 191a3114836SGerry Liu } 192a3114836SGerry Liu 1935d07b933Sdp78419 #define PAGE_GET_COLOR_SHIFT(szc, nszc) \ 1945d07b933Sdp78419 (hw_page_array[(nszc)].hp_shift - hw_page_array[(szc)].hp_shift) 1955d07b933Sdp78419 196ce8eb11aSdp78419 #define PAGE_CONVERT_COLOR(ncolor, szc, nszc) \ 197ce8eb11aSdp78419 ((ncolor) << PAGE_GET_COLOR_SHIFT((szc), (nszc))) 198ce8eb11aSdp78419 199ce8eb11aSdp78419 #define PFN_2_COLOR(pfn, szc, it) \ 2005d07b933Sdp78419 (((pfn) & page_colors_mask) >> \ 2017c478bd9Sstevel@tonic-gate (hw_page_array[szc].hp_shift - hw_page_array[0].hp_shift)) 2027c478bd9Sstevel@tonic-gate 2035d07b933Sdp78419 #define PNUM_SIZE(szc) \ 2045d07b933Sdp78419 (hw_page_array[(szc)].hp_pgcnt) 2055d07b933Sdp78419 #define PNUM_SHIFT(szc) \ 2065d07b933Sdp78419 (hw_page_array[(szc)].hp_shift - hw_page_array[0].hp_shift) 2075d07b933Sdp78419 #define PAGE_GET_SHIFT(szc) \ 2085d07b933Sdp78419 (hw_page_array[(szc)].hp_shift) 2095d07b933Sdp78419 #define PAGE_GET_PAGECOLORS(szc) \ 2105d07b933Sdp78419 (hw_page_array[(szc)].hp_colors) 2115d07b933Sdp78419 2125d07b933Sdp78419 /* 2135d07b933Sdp78419 * This macro calculates the next sequential pfn with the specified 2145d07b933Sdp78419 * color using color equivalency mask 2155d07b933Sdp78419 */ 216ce8eb11aSdp78419 #define PAGE_NEXT_PFN_FOR_COLOR(pfn, szc, color, ceq_mask, color_mask, it) \ 2175d07b933Sdp78419 { \ 2185d07b933Sdp78419 uint_t pfn_shift = PAGE_BSZS_SHIFT(szc); \ 2195d07b933Sdp78419 pfn_t spfn = pfn >> pfn_shift; \ 2205d07b933Sdp78419 pfn_t stride = (ceq_mask) + 1; \ 221a7c3ca36Sdp78419 ASSERT(((color) & ~(ceq_mask)) == 0); \ 2225d07b933Sdp78419 ASSERT((((ceq_mask) + 1) & (ceq_mask)) == 0); \ 2235d07b933Sdp78419 if (((spfn ^ (color)) & (ceq_mask)) == 0) { \ 2245d07b933Sdp78419 pfn += stride << pfn_shift; \ 2255d07b933Sdp78419 } else { \ 2265d07b933Sdp78419 pfn = (spfn & ~(pfn_t)(ceq_mask)) | (color); \ 2275d07b933Sdp78419 pfn = (pfn > spfn ? pfn : pfn + stride) << pfn_shift; \ 2285d07b933Sdp78419 } \ 2295d07b933Sdp78419 } 2305d07b933Sdp78419 2315d07b933Sdp78419 /* get the color equivalency mask for the next szc */ 2325d07b933Sdp78419 #define PAGE_GET_NSZ_MASK(szc, mask) \ 2335d07b933Sdp78419 ((mask) >> (PAGE_GET_SHIFT((szc) + 1) - PAGE_GET_SHIFT(szc))) 2345d07b933Sdp78419 2355d07b933Sdp78419 /* get the color of the next szc */ 2365d07b933Sdp78419 #define PAGE_GET_NSZ_COLOR(szc, color) \ 2375d07b933Sdp78419 ((color) >> (PAGE_GET_SHIFT((szc) + 1) - PAGE_GET_SHIFT(szc))) 2385d07b933Sdp78419 2395d07b933Sdp78419 /* Find the bin for the given page if it was of size szc */ 240ce8eb11aSdp78419 #define PP_2_BIN_SZC(pp, szc) (PFN_2_COLOR(pp->p_pagenum, szc, NULL)) 2415d07b933Sdp78419 242d94ffb28Sjmcp #define PP_2_BIN(pp) (PP_2_BIN_SZC(pp, pp->p_szc)) 2437c478bd9Sstevel@tonic-gate 2447c478bd9Sstevel@tonic-gate #define PP_2_MEM_NODE(pp) (PFN_2_MEM_NODE(pp->p_pagenum)) 245843e1988Sjohnlev #define PP_2_MTYPE(pp) (pfn_2_mtype(pp->p_pagenum)) 2467c478bd9Sstevel@tonic-gate #define PP_2_SZC(pp) (pp->p_szc) 2477c478bd9Sstevel@tonic-gate 2487c478bd9Sstevel@tonic-gate #define SZCPAGES(szc) (1 << PAGE_BSZS_SHIFT(szc)) 2497c478bd9Sstevel@tonic-gate #define PFN_BASE(pfnum, szc) (pfnum & ~(SZCPAGES(szc) - 1)) 2507c478bd9Sstevel@tonic-gate 251d94ffb28Sjmcp /* 252d94ffb28Sjmcp * this structure is used for walking free page lists 253d94ffb28Sjmcp * controls when to split large pages into smaller pages, 254d94ffb28Sjmcp * and when to coalesce smaller pages into larger pages 255d94ffb28Sjmcp */ 256d94ffb28Sjmcp typedef struct page_list_walker { 257d94ffb28Sjmcp uint_t plw_colors; /* num of colors for szc */ 258d94ffb28Sjmcp uint_t plw_color_mask; /* colors-1 */ 259d94ffb28Sjmcp uint_t plw_bin_step; /* next bin: 1 or 2 */ 260d94ffb28Sjmcp uint_t plw_count; /* loop count */ 261d94ffb28Sjmcp uint_t plw_bin0; /* starting bin */ 262d94ffb28Sjmcp uint_t plw_bin_marker; /* bin after initial jump */ 263d94ffb28Sjmcp uint_t plw_bin_split_prev; /* last bin we tried to split */ 264d94ffb28Sjmcp uint_t plw_do_split; /* set if OK to split */ 265d94ffb28Sjmcp uint_t plw_split_next; /* next bin to split */ 266d94ffb28Sjmcp uint_t plw_ceq_dif; /* number of different color groups */ 267d94ffb28Sjmcp /* to check */ 268d94ffb28Sjmcp uint_t plw_ceq_mask[MMU_PAGE_SIZES + 1]; /* color equiv mask */ 269d94ffb28Sjmcp uint_t plw_bins[MMU_PAGE_SIZES + 1]; /* num of bins */ 270d94ffb28Sjmcp } page_list_walker_t; 271d94ffb28Sjmcp 2725d07b933Sdp78419 void page_list_walk_init(uchar_t szc, uint_t flags, uint_t bin, 2735d07b933Sdp78419 int can_split, int use_ceq, page_list_walker_t *plw); 2745d07b933Sdp78419 2755d07b933Sdp78419 uint_t page_list_walk_next_bin(uchar_t szc, uint_t bin, 2765d07b933Sdp78419 page_list_walker_t *plw); 2775d07b933Sdp78419 278affbd3ccSkchow extern struct cpu cpus[]; 279affbd3ccSkchow #define CPU0 cpus 280affbd3ccSkchow 281843e1988Sjohnlev extern int mtype_init(vnode_t *, caddr_t, uint_t *, size_t); 282843e1988Sjohnlev #define MTYPE_INIT(mtype, vp, vaddr, flags, pgsz) \ 283843e1988Sjohnlev (mtype = mtype_init(vp, vaddr, &(flags), pgsz)) 2847c478bd9Sstevel@tonic-gate 2857c478bd9Sstevel@tonic-gate /* 2867c478bd9Sstevel@tonic-gate * macros to loop through the mtype range (page_get_mnode_{free,cache,any}list, 2877c478bd9Sstevel@tonic-gate * and page_get_contig_pages) 2887c478bd9Sstevel@tonic-gate * 2897c478bd9Sstevel@tonic-gate * MTYPE_START sets the initial mtype. -1 if the mtype range specified does 2907c478bd9Sstevel@tonic-gate * not contain mnode. 2917c478bd9Sstevel@tonic-gate * 2927c478bd9Sstevel@tonic-gate * MTYPE_NEXT sets the next mtype. -1 if there are no more valid 2937c478bd9Sstevel@tonic-gate * mtype in the range. 2947c478bd9Sstevel@tonic-gate */ 2957c478bd9Sstevel@tonic-gate 2967c478bd9Sstevel@tonic-gate #define MTYPE_START(mnode, mtype, flags) \ 2977c478bd9Sstevel@tonic-gate (mtype = mtype_func(mnode, mtype, flags)) 2987c478bd9Sstevel@tonic-gate 299affbd3ccSkchow #define MTYPE_NEXT(mnode, mtype, flags) { \ 300affbd3ccSkchow if (flags & PGI_MT_RANGE) { \ 301affbd3ccSkchow mtype = mtype_func(mnode, mtype, flags | PGI_MT_NEXT); \ 302affbd3ccSkchow } else { \ 303affbd3ccSkchow mtype = -1; \ 304affbd3ccSkchow } \ 305affbd3ccSkchow } 3067c478bd9Sstevel@tonic-gate 307843e1988Sjohnlev extern int mtype_pgr_init(int *, page_t *, int, pgcnt_t); 308843e1988Sjohnlev #define MTYPE_PGR_INIT(mtype, flags, pp, mnode, pgcnt) \ 309843e1988Sjohnlev (mtype = mtype_pgr_init(&flags, pp, mnode, pgcnt)) 3107c478bd9Sstevel@tonic-gate 311affbd3ccSkchow #define MNODE_PGCNT(mnode) mnode_pgcnt(mnode) 312affbd3ccSkchow 313843e1988Sjohnlev extern void mnodetype_2_pfn(int, int, pfn_t *, pfn_t *); 3147c478bd9Sstevel@tonic-gate #define MNODETYPE_2_PFN(mnode, mtype, pfnlo, pfnhi) \ 315843e1988Sjohnlev mnodetype_2_pfn(mnode, mtype, &pfnlo, &pfnhi) 3167c478bd9Sstevel@tonic-gate 317d94ffb28Sjmcp #define PC_BIN_MUTEX(mnode, bin, flags) ((flags & PG_FREE_LIST) ? \ 318d94ffb28Sjmcp &fpc_mutex[(bin) & (NPC_MUTEX - 1)][mnode] : \ 3197c478bd9Sstevel@tonic-gate &cpc_mutex[(bin) & (NPC_MUTEX - 1)][mnode]) 3207c478bd9Sstevel@tonic-gate 3217c478bd9Sstevel@tonic-gate #define FPC_MUTEX(mnode, i) (&fpc_mutex[i][mnode]) 3227c478bd9Sstevel@tonic-gate #define CPC_MUTEX(mnode, i) (&cpc_mutex[i][mnode]) 3237c478bd9Sstevel@tonic-gate 3247c478bd9Sstevel@tonic-gate #ifdef DEBUG 3257c478bd9Sstevel@tonic-gate #define CHK_LPG(pp, szc) chk_lpg(pp, szc) 3267c478bd9Sstevel@tonic-gate extern void chk_lpg(page_t *, uchar_t); 3277c478bd9Sstevel@tonic-gate #else 3287c478bd9Sstevel@tonic-gate #define CHK_LPG(pp, szc) 3297c478bd9Sstevel@tonic-gate #endif 3307c478bd9Sstevel@tonic-gate 3317c478bd9Sstevel@tonic-gate #define FULL_REGION_CNT(rg_szc) \ 3327c478bd9Sstevel@tonic-gate (LEVEL_SIZE(rg_szc) >> LEVEL_SHIFT(rg_szc - 1)) 3337c478bd9Sstevel@tonic-gate 3347c478bd9Sstevel@tonic-gate /* Return the leader for this mapping size */ 3357c478bd9Sstevel@tonic-gate #define PP_GROUPLEADER(pp, szc) \ 3367c478bd9Sstevel@tonic-gate (&(pp)[-(int)((pp)->p_pagenum & (SZCPAGES(szc)-1))]) 3377c478bd9Sstevel@tonic-gate 3387c478bd9Sstevel@tonic-gate /* Return the root page for this page based on p_szc */ 3397c478bd9Sstevel@tonic-gate #define PP_PAGEROOT(pp) ((pp)->p_szc == 0 ? (pp) : \ 3407c478bd9Sstevel@tonic-gate PP_GROUPLEADER((pp), (pp)->p_szc)) 3417c478bd9Sstevel@tonic-gate 3427c478bd9Sstevel@tonic-gate /* 3437c478bd9Sstevel@tonic-gate * The counter base must be per page_counter element to prevent 3447c478bd9Sstevel@tonic-gate * races when re-indexing, and the base page size element should 3457c478bd9Sstevel@tonic-gate * be aligned on a boundary of the given region size. 3467c478bd9Sstevel@tonic-gate * 3477c478bd9Sstevel@tonic-gate * We also round up the number of pages spanned by the counters 3487c478bd9Sstevel@tonic-gate * for a given region to PC_BASE_ALIGN in certain situations to simplify 3497c478bd9Sstevel@tonic-gate * the coding for some non-performance critical routines. 3507c478bd9Sstevel@tonic-gate */ 3517c478bd9Sstevel@tonic-gate 3527c478bd9Sstevel@tonic-gate #define PC_BASE_ALIGN ((pfn_t)1 << PAGE_BSZS_SHIFT(MMU_PAGE_SIZES-1)) 3537c478bd9Sstevel@tonic-gate #define PC_BASE_ALIGN_MASK (PC_BASE_ALIGN - 1) 3547c478bd9Sstevel@tonic-gate 3557c478bd9Sstevel@tonic-gate /* 3567c478bd9Sstevel@tonic-gate * cpu/mmu-dependent vm variables 3577c478bd9Sstevel@tonic-gate */ 3587c478bd9Sstevel@tonic-gate extern uint_t mmu_page_sizes; 3597c478bd9Sstevel@tonic-gate extern uint_t mmu_exported_page_sizes; 36002bc52beSkchow /* 36102bc52beSkchow * page sizes that legacy applications can see via getpagesizes(3c). 36202bc52beSkchow * Used to prevent legacy applications from inadvertantly using the 36302bc52beSkchow * 'new' large pagesizes (1g and above). 36402bc52beSkchow */ 36502bc52beSkchow extern uint_t mmu_legacy_page_sizes; 3667c478bd9Sstevel@tonic-gate 3677c478bd9Sstevel@tonic-gate /* For x86, userszc is the same as the kernel's szc */ 3687c478bd9Sstevel@tonic-gate #define USERSZC_2_SZC(userszc) (userszc) 3697c478bd9Sstevel@tonic-gate #define SZC_2_USERSZC(szc) (szc) 3707c478bd9Sstevel@tonic-gate 3717c478bd9Sstevel@tonic-gate /* 3727c478bd9Sstevel@tonic-gate * for hw_page_map_t, sized to hold the ratio of large page to base 3737c478bd9Sstevel@tonic-gate * pagesize (1024 max) 3747c478bd9Sstevel@tonic-gate */ 3757c478bd9Sstevel@tonic-gate typedef short hpmctr_t; 3767c478bd9Sstevel@tonic-gate 3777c478bd9Sstevel@tonic-gate /* 3787c478bd9Sstevel@tonic-gate * get the setsize of the current cpu - assume homogenous for x86 3797c478bd9Sstevel@tonic-gate */ 3807c478bd9Sstevel@tonic-gate extern int l2cache_sz, l2cache_linesz, l2cache_assoc; 3817c478bd9Sstevel@tonic-gate 3827c478bd9Sstevel@tonic-gate #define L2CACHE_ALIGN l2cache_linesz 383affbd3ccSkchow #define L2CACHE_ALIGN_MAX 64 3847c478bd9Sstevel@tonic-gate #define CPUSETSIZE() \ 3857c478bd9Sstevel@tonic-gate (l2cache_assoc ? (l2cache_sz / l2cache_assoc) : MMU_PAGESIZE) 3867c478bd9Sstevel@tonic-gate 3877c478bd9Sstevel@tonic-gate /* 3887c478bd9Sstevel@tonic-gate * Return the log2(pagesize(szc) / MMU_PAGESIZE) --- or the shift count 3897c478bd9Sstevel@tonic-gate * for the number of base pages in this pagesize 3907c478bd9Sstevel@tonic-gate */ 3917c478bd9Sstevel@tonic-gate #define PAGE_BSZS_SHIFT(szc) (LEVEL_SHIFT(szc) - MMU_PAGESHIFT) 3927c478bd9Sstevel@tonic-gate 3937c478bd9Sstevel@tonic-gate /* 3947c478bd9Sstevel@tonic-gate * Internal PG_ flags. 3957c478bd9Sstevel@tonic-gate */ 3967c478bd9Sstevel@tonic-gate #define PGI_RELOCONLY 0x010000 /* opposite of PG_NORELOC */ 3977c478bd9Sstevel@tonic-gate #define PGI_NOCAGE 0x020000 /* cage is disabled */ 3987c478bd9Sstevel@tonic-gate #define PGI_PGCPHIPRI 0x040000 /* page_get_contig_page pri alloc */ 3997c478bd9Sstevel@tonic-gate #define PGI_PGCPSZC0 0x080000 /* relocate base pagesize page */ 4007c478bd9Sstevel@tonic-gate 4017c478bd9Sstevel@tonic-gate /* 4027c478bd9Sstevel@tonic-gate * PGI range flags - should not overlap PGI flags 4037c478bd9Sstevel@tonic-gate */ 4047c478bd9Sstevel@tonic-gate #define PGI_MT_RANGE0 0x1000000 /* mtype range to 0 */ 40507ad560dSkchow #define PGI_MT_RANGE16M 0x2000000 /* mtype range to 16m */ 40607ad560dSkchow #define PGI_MT_RANGE4G 0x4000000 /* mtype range to 4g */ 40707ad560dSkchow #define PGI_MT_NEXT 0x8000000 /* get next mtype */ 40807ad560dSkchow #define PGI_MT_RANGE (PGI_MT_RANGE0 | PGI_MT_RANGE16M | PGI_MT_RANGE4G) 4097c478bd9Sstevel@tonic-gate 410843e1988Sjohnlev 4117c478bd9Sstevel@tonic-gate /* 412ec25b48fSsusans * Maximum and default values for user heap, stack, private and shared 413ec25b48fSsusans * anonymous memory, and user text and initialized data. 414ec25b48fSsusans * Used by map_pgsz*() routines. 415ec25b48fSsusans */ 416ec25b48fSsusans extern size_t max_uheap_lpsize; 417ec25b48fSsusans extern size_t default_uheap_lpsize; 418ec25b48fSsusans extern size_t max_ustack_lpsize; 419ec25b48fSsusans extern size_t default_ustack_lpsize; 420ec25b48fSsusans extern size_t max_privmap_lpsize; 421ec25b48fSsusans extern size_t max_uidata_lpsize; 422ec25b48fSsusans extern size_t max_utext_lpsize; 423ec25b48fSsusans extern size_t max_shm_lpsize; 424ec25b48fSsusans extern size_t mcntl0_lpsize; 425ec25b48fSsusans 426ec25b48fSsusans /* 427ec25b48fSsusans * Sanity control. Don't use large pages regardless of user 428ec25b48fSsusans * settings if there's less than priv or shm_lpg_min_physmem memory installed. 429ec25b48fSsusans * The units for this variable are 8K pages. 430ec25b48fSsusans */ 431ec25b48fSsusans extern pgcnt_t privm_lpg_min_physmem; 432ec25b48fSsusans extern pgcnt_t shm_lpg_min_physmem; 433ec25b48fSsusans 434ec25b48fSsusans /* 4357c478bd9Sstevel@tonic-gate * hash as and addr to get a bin. 4367c478bd9Sstevel@tonic-gate */ 4377c478bd9Sstevel@tonic-gate 438d94ffb28Sjmcp #define AS_2_BIN(as, seg, vp, addr, bin, szc) \ 4395d07b933Sdp78419 bin = (((((uintptr_t)(addr) >> PAGESHIFT) + ((uintptr_t)(as) >> 4)) \ 4405d07b933Sdp78419 & page_colors_mask) >> \ 4415d07b933Sdp78419 (hw_page_array[szc].hp_shift - hw_page_array[0].hp_shift)) 4427c478bd9Sstevel@tonic-gate 4437c478bd9Sstevel@tonic-gate /* 444affbd3ccSkchow * cpu private vm data - accessed thru CPU->cpu_vm_data 445affbd3ccSkchow * vc_pnum_memseg: tracks last memseg visited in page_numtopp_nolock() 446affbd3ccSkchow * vc_pnext_memseg: tracks last memseg visited in page_nextn() 447affbd3ccSkchow * vc_kmptr: orignal unaligned kmem pointer for this vm_cpu_data_t 4486061ce8aSkchow * vc_kmsize: orignal kmem size for this vm_cpu_data_t 449affbd3ccSkchow */ 450affbd3ccSkchow 451affbd3ccSkchow typedef struct { 452affbd3ccSkchow struct memseg *vc_pnum_memseg; 453affbd3ccSkchow struct memseg *vc_pnext_memseg; 454affbd3ccSkchow void *vc_kmptr; 4556061ce8aSkchow size_t vc_kmsize; 456affbd3ccSkchow } vm_cpu_data_t; 457affbd3ccSkchow 458affbd3ccSkchow /* allocation size to ensure vm_cpu_data_t resides in its own cache line */ 459affbd3ccSkchow #define VM_CPU_DATA_PADSIZE \ 460affbd3ccSkchow (P2ROUNDUP(sizeof (vm_cpu_data_t), L2CACHE_ALIGN_MAX)) 461affbd3ccSkchow 462affbd3ccSkchow /* 4637c478bd9Sstevel@tonic-gate * When a bin is empty, and we can't satisfy a color request correctly, 4647c478bd9Sstevel@tonic-gate * we scan. If we assume that the programs have reasonable spatial 4657c478bd9Sstevel@tonic-gate * behavior, then it will not be a good idea to use the adjacent color. 4667c478bd9Sstevel@tonic-gate * Using the adjacent color would result in virtually adjacent addresses 4677c478bd9Sstevel@tonic-gate * mapping into the same spot in the cache. So, if we stumble across 4687c478bd9Sstevel@tonic-gate * an empty bin, skip a bunch before looking. After the first skip, 4697c478bd9Sstevel@tonic-gate * then just look one bin at a time so we don't miss our cache on 4707c478bd9Sstevel@tonic-gate * every look. Be sure to check every bin. Page_create() will panic 4717c478bd9Sstevel@tonic-gate * if we miss a page. 4727c478bd9Sstevel@tonic-gate * 4737c478bd9Sstevel@tonic-gate * This also explains the `<=' in the for loops in both page_get_freelist() 4747c478bd9Sstevel@tonic-gate * and page_get_cachelist(). Since we checked the target bin, skipped 4757c478bd9Sstevel@tonic-gate * a bunch, then continued one a time, we wind up checking the target bin 4767c478bd9Sstevel@tonic-gate * twice to make sure we get all of them bins. 4777c478bd9Sstevel@tonic-gate */ 4787c478bd9Sstevel@tonic-gate #define BIN_STEP 19 4797c478bd9Sstevel@tonic-gate 4807c478bd9Sstevel@tonic-gate #ifdef VM_STATS 4817c478bd9Sstevel@tonic-gate struct vmm_vmstats_str { 482d94ffb28Sjmcp ulong_t pgf_alloc[MMU_PAGE_SIZES]; /* page_get_freelist */ 483d94ffb28Sjmcp ulong_t pgf_allocok[MMU_PAGE_SIZES]; 484d94ffb28Sjmcp ulong_t pgf_allocokrem[MMU_PAGE_SIZES]; 485d94ffb28Sjmcp ulong_t pgf_allocfailed[MMU_PAGE_SIZES]; 4867c478bd9Sstevel@tonic-gate ulong_t pgf_allocdeferred; 487d94ffb28Sjmcp ulong_t pgf_allocretry[MMU_PAGE_SIZES]; 488affbd3ccSkchow ulong_t pgc_alloc; /* page_get_cachelist */ 4897c478bd9Sstevel@tonic-gate ulong_t pgc_allocok; 4907c478bd9Sstevel@tonic-gate ulong_t pgc_allocokrem; 4917c478bd9Sstevel@tonic-gate ulong_t pgc_allocokdeferred; 4927c478bd9Sstevel@tonic-gate ulong_t pgc_allocfailed; 493affbd3ccSkchow ulong_t pgcp_alloc[MMU_PAGE_SIZES]; /* page_get_contig_pages */ 4947c478bd9Sstevel@tonic-gate ulong_t pgcp_allocfailed[MMU_PAGE_SIZES]; 4957c478bd9Sstevel@tonic-gate ulong_t pgcp_allocempty[MMU_PAGE_SIZES]; 4967c478bd9Sstevel@tonic-gate ulong_t pgcp_allocok[MMU_PAGE_SIZES]; 497affbd3ccSkchow ulong_t ptcp[MMU_PAGE_SIZES]; /* page_trylock_contig_pages */ 4987c478bd9Sstevel@tonic-gate ulong_t ptcpfreethresh[MMU_PAGE_SIZES]; 4997c478bd9Sstevel@tonic-gate ulong_t ptcpfailexcl[MMU_PAGE_SIZES]; 5007c478bd9Sstevel@tonic-gate ulong_t ptcpfailszc[MMU_PAGE_SIZES]; 5017c478bd9Sstevel@tonic-gate ulong_t ptcpfailcage[MMU_PAGE_SIZES]; 5027c478bd9Sstevel@tonic-gate ulong_t ptcpok[MMU_PAGE_SIZES]; 503affbd3ccSkchow ulong_t pgmf_alloc[MMU_PAGE_SIZES]; /* page_get_mnode_freelist */ 5047c478bd9Sstevel@tonic-gate ulong_t pgmf_allocfailed[MMU_PAGE_SIZES]; 5057c478bd9Sstevel@tonic-gate ulong_t pgmf_allocempty[MMU_PAGE_SIZES]; 5067c478bd9Sstevel@tonic-gate ulong_t pgmf_allocok[MMU_PAGE_SIZES]; 507affbd3ccSkchow ulong_t pgmc_alloc; /* page_get_mnode_cachelist */ 5087c478bd9Sstevel@tonic-gate ulong_t pgmc_allocfailed; 5097c478bd9Sstevel@tonic-gate ulong_t pgmc_allocempty; 5107c478bd9Sstevel@tonic-gate ulong_t pgmc_allocok; 511affbd3ccSkchow ulong_t pladd_free[MMU_PAGE_SIZES]; /* page_list_add/sub */ 512affbd3ccSkchow ulong_t plsub_free[MMU_PAGE_SIZES]; 513affbd3ccSkchow ulong_t pladd_cache; 514affbd3ccSkchow ulong_t plsub_cache; 515affbd3ccSkchow ulong_t plsubpages_szcbig; 516affbd3ccSkchow ulong_t plsubpages_szc0; 5175d07b933Sdp78419 ulong_t pfs_req[MMU_PAGE_SIZES]; /* page_freelist_split */ 5185d07b933Sdp78419 ulong_t pfs_demote[MMU_PAGE_SIZES]; 5195d07b933Sdp78419 ulong_t pfc_coalok[MMU_PAGE_SIZES][MAX_MNODE_MRANGES]; 520affbd3ccSkchow ulong_t ppr_reloc[MMU_PAGE_SIZES]; /* page_relocate */ 5217c478bd9Sstevel@tonic-gate ulong_t ppr_relocnoroot[MMU_PAGE_SIZES]; 5227c478bd9Sstevel@tonic-gate ulong_t ppr_reloc_replnoroot[MMU_PAGE_SIZES]; 5237c478bd9Sstevel@tonic-gate ulong_t ppr_relocnolock[MMU_PAGE_SIZES]; 5247c478bd9Sstevel@tonic-gate ulong_t ppr_relocnomem[MMU_PAGE_SIZES]; 5257c478bd9Sstevel@tonic-gate ulong_t ppr_relocok[MMU_PAGE_SIZES]; 5268b464eb8Smec ulong_t ppr_copyfail; 5275d07b933Sdp78419 /* page coalesce counter */ 5285d07b933Sdp78419 ulong_t page_ctrs_coalesce[MMU_PAGE_SIZES][MAX_MNODE_MRANGES]; 5295d07b933Sdp78419 /* candidates useful */ 5305d07b933Sdp78419 ulong_t page_ctrs_cands_skip[MMU_PAGE_SIZES][MAX_MNODE_MRANGES]; 5315d07b933Sdp78419 /* ctrs changed after locking */ 5325d07b933Sdp78419 ulong_t page_ctrs_changed[MMU_PAGE_SIZES][MAX_MNODE_MRANGES]; 5335d07b933Sdp78419 /* page_freelist_coalesce failed */ 5345d07b933Sdp78419 ulong_t page_ctrs_failed[MMU_PAGE_SIZES][MAX_MNODE_MRANGES]; 5357c478bd9Sstevel@tonic-gate ulong_t page_ctrs_coalesce_all; /* page coalesce all counter */ 5367c478bd9Sstevel@tonic-gate ulong_t page_ctrs_cands_skip_all; /* candidates useful for all func */ 5377c478bd9Sstevel@tonic-gate ulong_t restrict4gcnt; 53807ad560dSkchow ulong_t unrestrict16mcnt; /* non-DMA 16m allocs allowed */ 539a985e578Skchow ulong_t pgpanicalloc; /* PG_PANIC allocation */ 54078b03d3aSkchow ulong_t pcf_deny[MMU_PAGE_SIZES]; /* page_chk_freelist */ 54178b03d3aSkchow ulong_t pcf_allow[MMU_PAGE_SIZES]; 5427c478bd9Sstevel@tonic-gate }; 5437c478bd9Sstevel@tonic-gate extern struct vmm_vmstats_str vmm_vmstats; 5447c478bd9Sstevel@tonic-gate #endif /* VM_STATS */ 5457c478bd9Sstevel@tonic-gate 5467c478bd9Sstevel@tonic-gate extern size_t page_ctrs_sz(void); 5477c478bd9Sstevel@tonic-gate extern caddr_t page_ctrs_alloc(caddr_t); 548affbd3ccSkchow extern void page_ctr_sub(int, int, page_t *, int); 5495d07b933Sdp78419 extern page_t *page_freelist_split(uchar_t, 55019397407SSherry Moore uint_t, int, int, pfn_t, pfn_t, page_list_walker_t *); 5515d07b933Sdp78419 extern page_t *page_freelist_coalesce(int, uchar_t, uint_t, uint_t, int, 5525d07b933Sdp78419 pfn_t); 553a3114836SGerry Liu extern void page_freelist_coalesce_all(int); 5547c478bd9Sstevel@tonic-gate extern uint_t page_get_pagecolors(uint_t); 555d2b85481Srscott extern void pfnzero(pfn_t, uint_t, uint_t); 5567c478bd9Sstevel@tonic-gate 5577c478bd9Sstevel@tonic-gate #ifdef __cplusplus 5587c478bd9Sstevel@tonic-gate } 5597c478bd9Sstevel@tonic-gate #endif 5607c478bd9Sstevel@tonic-gate 5617c478bd9Sstevel@tonic-gate #endif /* _VM_DEP_H */ 562