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 /* 22b52a336eSPavel Tatashin * Copyright (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved. 23d3b5f563SJohn Levon * Copyright 2019 Joyent, Inc. 247c478bd9Sstevel@tonic-gate */ 25a3114836SGerry Liu /* 26a3114836SGerry Liu * Copyright (c) 2010, Intel Corporation. 27a3114836SGerry Liu * All rights reserved. 28a3114836SGerry Liu */ 297c478bd9Sstevel@tonic-gate 307c478bd9Sstevel@tonic-gate /* 317c478bd9Sstevel@tonic-gate * UNIX machine dependent virtual memory support. 327c478bd9Sstevel@tonic-gate */ 337c478bd9Sstevel@tonic-gate 347c478bd9Sstevel@tonic-gate #ifndef _VM_DEP_H 357c478bd9Sstevel@tonic-gate #define _VM_DEP_H 367c478bd9Sstevel@tonic-gate 377c478bd9Sstevel@tonic-gate 387c478bd9Sstevel@tonic-gate #ifdef __cplusplus 397c478bd9Sstevel@tonic-gate extern "C" { 407c478bd9Sstevel@tonic-gate #endif 417c478bd9Sstevel@tonic-gate 427c478bd9Sstevel@tonic-gate #include <sys/clock.h> 437c478bd9Sstevel@tonic-gate #include <vm/hat_pte.h> 44ae115bc7Smrj #include <sys/param.h> 45ce8eb11aSdp #include <sys/memnode.h> 467c478bd9Sstevel@tonic-gate 477c478bd9Sstevel@tonic-gate /* 48843e1988Sjohnlev * WARNING: vm_dep.h is included by files in common. 497c478bd9Sstevel@tonic-gate */ 507c478bd9Sstevel@tonic-gate 517c478bd9Sstevel@tonic-gate #define GETTICK() tsc_read() 5206fb6a36Sdv /* 5306fb6a36Sdv * Do not use this function for obtaining clock tick. This 5406fb6a36Sdv * is called by callers who do not need to have a guarenteed 5506fb6a36Sdv * correct tick value. The proper routine to use is tsc_read(). 5606fb6a36Sdv */ 577c478bd9Sstevel@tonic-gate 58b52a336eSPavel Tatashin extern u_longlong_t randtick(); 59843e1988Sjohnlev extern uint_t page_create_update_flags_x86(uint_t); 607c478bd9Sstevel@tonic-gate 61843e1988Sjohnlev extern size_t plcnt_sz(size_t); 62843e1988Sjohnlev #define PLCNT_SZ(ctrs_sz) (ctrs_sz = plcnt_sz(ctrs_sz)) 637c478bd9Sstevel@tonic-gate 64843e1988Sjohnlev extern caddr_t plcnt_init(caddr_t); 65843e1988Sjohnlev #define PLCNT_INIT(addr) (addr = plcnt_init(addr)) 6607ad560dSkchow 67843e1988Sjohnlev extern void plcnt_inc_dec(page_t *, int, int, long, int); 68843e1988Sjohnlev #define PLCNT_INCR(pp, mnode, mtype, szc, flags) \ 69843e1988Sjohnlev plcnt_inc_dec(pp, mtype, szc, 1l << PAGE_BSZS_SHIFT(szc), flags) 70843e1988Sjohnlev #define PLCNT_DECR(pp, mnode, mtype, szc, flags) \ 71567fb7baSToomas Soome plcnt_inc_dec(pp, mtype, szc, \ 72567fb7baSToomas Soome (long)(ULONG_MAX << PAGE_BSZS_SHIFT(szc)), flags) 737c478bd9Sstevel@tonic-gate 74affbd3ccSkchow /* 75843e1988Sjohnlev * macro to update page list max counts. no-op on x86. 76affbd3ccSkchow */ 77e21bae1bSkchow #define PLCNT_XFER_NORELOC(pp) 78e21bae1bSkchow 79a3114836SGerry Liu /* 80a3114836SGerry Liu * macro to modify the page list max counts when memory is added to 81a3114836SGerry Liu * the page lists during startup (add_physmem) or during a DR operation 82a3114836SGerry Liu * when memory is added (kphysm_add_memory_dynamic) or deleted 83a3114836SGerry Liu * (kphysm_del_cleanup). 84a3114836SGerry Liu */ 85a3114836SGerry Liu #define PLCNT_MODIFY_MAX(pfn, cnt) mtype_modify_max(pfn, cnt) 86a3114836SGerry Liu 87843e1988Sjohnlev extern int memrange_num(pfn_t); 88843e1988Sjohnlev extern int pfn_2_mtype(pfn_t); 89843e1988Sjohnlev extern int mtype_func(int, int, uint_t); 90843e1988Sjohnlev extern void mtype_modify_max(pfn_t, long); 91843e1988Sjohnlev extern int mnode_pgcnt(int); 92843e1988Sjohnlev extern int mnode_range_cnt(int); 937c478bd9Sstevel@tonic-gate 9407ad560dSkchow /* 95843e1988Sjohnlev * candidate counters in vm_pagelist.c are indexed by color and range 9607ad560dSkchow */ 97a3114836SGerry Liu #define NUM_MEM_RANGES 4 /* memory range types */ 98843e1988Sjohnlev #define MAX_MNODE_MRANGES NUM_MEM_RANGES 99843e1988Sjohnlev #define MNODE_RANGE_CNT(mnode) mnode_range_cnt(mnode) 100843e1988Sjohnlev #define MNODE_MAX_MRANGE(mnode) memrange_num(mem_node_config[mnode].physbase) 1017c478bd9Sstevel@tonic-gate 1025d07b933Sdp /* 103843e1988Sjohnlev * This was really badly defined, it implicitly uses mnode_maxmrange[] 104843e1988Sjohnlev * which is a static in vm_pagelist.c 1055d07b933Sdp */ 106843e1988Sjohnlev extern int mtype_2_mrange(int); 1075d07b933Sdp #define MTYPE_2_MRANGE(mnode, mtype) \ 108843e1988Sjohnlev (mnode_maxmrange[mnode] - mtype_2_mrange(mtype)) 1095d07b933Sdp 1107c478bd9Sstevel@tonic-gate /* 111d94ffb28Sjmcp * Per page size free lists. Allocated dynamically. 1127c478bd9Sstevel@tonic-gate * dimensions [mtype][mmu_page_sizes][colors] 1137c478bd9Sstevel@tonic-gate * 1147c478bd9Sstevel@tonic-gate * mtype specifies a physical memory range with a unique mnode. 1157c478bd9Sstevel@tonic-gate */ 1167c478bd9Sstevel@tonic-gate 117d94ffb28Sjmcp extern page_t ****page_freelists; 1187c478bd9Sstevel@tonic-gate 119d94ffb28Sjmcp #define PAGE_FREELISTS(mnode, szc, color, mtype) \ 120d94ffb28Sjmcp (*(page_freelists[mtype][szc] + (color))) 1217c478bd9Sstevel@tonic-gate 1227c478bd9Sstevel@tonic-gate /* 1237c478bd9Sstevel@tonic-gate * For now there is only a single size cache list. Allocated dynamically. 1247c478bd9Sstevel@tonic-gate * dimensions [mtype][colors] 1257c478bd9Sstevel@tonic-gate * 1267c478bd9Sstevel@tonic-gate * mtype specifies a physical memory range with a unique mnode. 1277c478bd9Sstevel@tonic-gate */ 1287c478bd9Sstevel@tonic-gate extern page_t ***page_cachelists; 1297c478bd9Sstevel@tonic-gate 130*584b574aSToomas Soome #define PAGE_CACHELISTS(mnode, color, mtype) \ 1317c478bd9Sstevel@tonic-gate (*(page_cachelists[mtype] + (color))) 1327c478bd9Sstevel@tonic-gate 1337c478bd9Sstevel@tonic-gate /* 134d94ffb28Sjmcp * There are mutexes for both the page freelist 1357c478bd9Sstevel@tonic-gate * and the page cachelist. We want enough locks to make contention 1367c478bd9Sstevel@tonic-gate * reasonable, but not too many -- otherwise page_freelist_lock() gets 1377c478bd9Sstevel@tonic-gate * so expensive that it becomes the bottleneck! 1387c478bd9Sstevel@tonic-gate */ 1397c478bd9Sstevel@tonic-gate 1407c478bd9Sstevel@tonic-gate #define NPC_MUTEX 16 1417c478bd9Sstevel@tonic-gate 1427c478bd9Sstevel@tonic-gate extern kmutex_t *fpc_mutex[NPC_MUTEX]; 1437c478bd9Sstevel@tonic-gate extern kmutex_t *cpc_mutex[NPC_MUTEX]; 1447c478bd9Sstevel@tonic-gate 145d94ffb28Sjmcp extern page_t *page_get_mnode_freelist(int, uint_t, int, uchar_t, uint_t); 1467c478bd9Sstevel@tonic-gate extern page_t *page_get_mnode_cachelist(uint_t, uint_t, int, int); 1477c478bd9Sstevel@tonic-gate 148ce8eb11aSdp /* mem node iterator is not used on x86 */ 149ce8eb11aSdp #define MEM_NODE_ITERATOR_DECL(it) 150b779d3e0Sdp #define MEM_NODE_ITERATOR_INIT(pfn, mnode, szc, it) 151ce8eb11aSdp 152ce8eb11aSdp /* 153ce8eb11aSdp * interleaved_mnodes mode is never set on x86, therefore, 154ce8eb11aSdp * simply return the limits of the given mnode, which then 155ce8eb11aSdp * determines the length of hpm_counters array for the mnode. 156ce8eb11aSdp */ 157*584b574aSToomas Soome #define HPM_COUNTERS_LIMITS(mnode, physbase, physmax, first) \ 158ce8eb11aSdp { \ 159ce8eb11aSdp (physbase) = mem_node_config[(mnode)].physbase; \ 160ce8eb11aSdp (physmax) = mem_node_config[(mnode)].physmax; \ 161ce8eb11aSdp (first) = (mnode); \ 162ce8eb11aSdp } 163ce8eb11aSdp 164ce8eb11aSdp #define PAGE_CTRS_WRITE_LOCK(mnode) \ 165ce8eb11aSdp { \ 166ce8eb11aSdp rw_enter(&page_ctrs_rwlock[(mnode)], RW_WRITER);\ 167ce8eb11aSdp page_freelist_lock(mnode); \ 168ce8eb11aSdp } 169ce8eb11aSdp 170ce8eb11aSdp #define PAGE_CTRS_WRITE_UNLOCK(mnode) \ 171ce8eb11aSdp { \ 172ce8eb11aSdp page_freelist_unlock(mnode); \ 173ce8eb11aSdp rw_exit(&page_ctrs_rwlock[(mnode)]); \ 174ce8eb11aSdp } 175ce8eb11aSdp 176a3114836SGerry Liu /* 177a3114836SGerry Liu * macro to call page_ctrs_adjust() when memory is added 178a3114836SGerry Liu * during a DR operation. 179a3114836SGerry Liu */ 180a3114836SGerry Liu #define PAGE_CTRS_ADJUST(pfn, cnt, rv) { \ 181a3114836SGerry Liu spgcnt_t _cnt = (spgcnt_t)(cnt); \ 182a3114836SGerry Liu int _mn; \ 183a3114836SGerry Liu pgcnt_t _np; \ 184a3114836SGerry Liu pfn_t _pfn = (pfn); \ 185a3114836SGerry Liu pfn_t _endpfn = _pfn + _cnt; \ 186*584b574aSToomas Soome rv = 0; \ 187a3114836SGerry Liu while (_pfn < _endpfn) { \ 188a3114836SGerry Liu _mn = PFN_2_MEM_NODE(_pfn); \ 189a3114836SGerry Liu _np = MIN(_endpfn, mem_node_config[_mn].physmax + 1) - _pfn; \ 190a3114836SGerry Liu _pfn += _np; \ 191a3114836SGerry Liu if ((rv = page_ctrs_adjust(_mn)) != 0) \ 192a3114836SGerry Liu break; \ 193a3114836SGerry Liu } \ 194a3114836SGerry Liu } 195a3114836SGerry Liu 196ce8eb11aSdp #define PAGE_GET_COLOR_SHIFT(szc, nszc) \ 1975d07b933Sdp (hw_page_array[(nszc)].hp_shift - hw_page_array[(szc)].hp_shift) 1985d07b933Sdp 199ce8eb11aSdp #define PAGE_CONVERT_COLOR(ncolor, szc, nszc) \ 200ce8eb11aSdp ((ncolor) << PAGE_GET_COLOR_SHIFT((szc), (nszc))) 201ce8eb11aSdp 202ce8eb11aSdp #define PFN_2_COLOR(pfn, szc, it) \ 2035d07b933Sdp (((pfn) & page_colors_mask) >> \ 2047c478bd9Sstevel@tonic-gate (hw_page_array[szc].hp_shift - hw_page_array[0].hp_shift)) 2057c478bd9Sstevel@tonic-gate 2065d07b933Sdp #define PNUM_SIZE(szc) \ 2075d07b933Sdp (hw_page_array[(szc)].hp_pgcnt) 2085d07b933Sdp #define PNUM_SHIFT(szc) \ 2095d07b933Sdp (hw_page_array[(szc)].hp_shift - hw_page_array[0].hp_shift) 2105d07b933Sdp #define PAGE_GET_SHIFT(szc) \ 2115d07b933Sdp (hw_page_array[(szc)].hp_shift) 2125d07b933Sdp #define PAGE_GET_PAGECOLORS(szc) \ 2135d07b933Sdp (hw_page_array[(szc)].hp_colors) 2145d07b933Sdp 2155d07b933Sdp /* 2165d07b933Sdp * This macro calculates the next sequential pfn with the specified 2175d07b933Sdp * color using color equivalency mask 2185d07b933Sdp */ 219ce8eb11aSdp #define PAGE_NEXT_PFN_FOR_COLOR(pfn, szc, color, ceq_mask, color_mask, it) \ 2205d07b933Sdp { \ 2215d07b933Sdp uint_t pfn_shift = PAGE_BSZS_SHIFT(szc); \ 2225d07b933Sdp pfn_t spfn = pfn >> pfn_shift; \ 2235d07b933Sdp pfn_t stride = (ceq_mask) + 1; \ 224a7c3ca36Sdp ASSERT(((color) & ~(ceq_mask)) == 0); \ 2255d07b933Sdp ASSERT((((ceq_mask) + 1) & (ceq_mask)) == 0); \ 2265d07b933Sdp if (((spfn ^ (color)) & (ceq_mask)) == 0) { \ 2275d07b933Sdp pfn += stride << pfn_shift; \ 2285d07b933Sdp } else { \ 2295d07b933Sdp pfn = (spfn & ~(pfn_t)(ceq_mask)) | (color); \ 2305d07b933Sdp pfn = (pfn > spfn ? pfn : pfn + stride) << pfn_shift; \ 2315d07b933Sdp } \ 2325d07b933Sdp } 2335d07b933Sdp 2345d07b933Sdp /* get the color equivalency mask for the next szc */ 2355d07b933Sdp #define PAGE_GET_NSZ_MASK(szc, mask) \ 2365d07b933Sdp ((mask) >> (PAGE_GET_SHIFT((szc) + 1) - PAGE_GET_SHIFT(szc))) 2375d07b933Sdp 2385d07b933Sdp /* get the color of the next szc */ 2395d07b933Sdp #define PAGE_GET_NSZ_COLOR(szc, color) \ 2405d07b933Sdp ((color) >> (PAGE_GET_SHIFT((szc) + 1) - PAGE_GET_SHIFT(szc))) 2415d07b933Sdp 2425d07b933Sdp /* Find the bin for the given page if it was of size szc */ 243ce8eb11aSdp #define PP_2_BIN_SZC(pp, szc) (PFN_2_COLOR(pp->p_pagenum, szc, NULL)) 2445d07b933Sdp 245d94ffb28Sjmcp #define PP_2_BIN(pp) (PP_2_BIN_SZC(pp, pp->p_szc)) 2467c478bd9Sstevel@tonic-gate 2477c478bd9Sstevel@tonic-gate #define PP_2_MEM_NODE(pp) (PFN_2_MEM_NODE(pp->p_pagenum)) 248843e1988Sjohnlev #define PP_2_MTYPE(pp) (pfn_2_mtype(pp->p_pagenum)) 2497c478bd9Sstevel@tonic-gate #define PP_2_SZC(pp) (pp->p_szc) 2507c478bd9Sstevel@tonic-gate 2517c478bd9Sstevel@tonic-gate #define SZCPAGES(szc) (1 << PAGE_BSZS_SHIFT(szc)) 2527c478bd9Sstevel@tonic-gate #define PFN_BASE(pfnum, szc) (pfnum & ~(SZCPAGES(szc) - 1)) 2537c478bd9Sstevel@tonic-gate 254d94ffb28Sjmcp /* 255d94ffb28Sjmcp * this structure is used for walking free page lists 256d94ffb28Sjmcp * controls when to split large pages into smaller pages, 257d94ffb28Sjmcp * and when to coalesce smaller pages into larger pages 258d94ffb28Sjmcp */ 259d94ffb28Sjmcp typedef struct page_list_walker { 260d94ffb28Sjmcp uint_t plw_colors; /* num of colors for szc */ 261d94ffb28Sjmcp uint_t plw_color_mask; /* colors-1 */ 262d94ffb28Sjmcp uint_t plw_bin_step; /* next bin: 1 or 2 */ 263d94ffb28Sjmcp uint_t plw_count; /* loop count */ 264d94ffb28Sjmcp uint_t plw_bin0; /* starting bin */ 265d94ffb28Sjmcp uint_t plw_bin_marker; /* bin after initial jump */ 266d94ffb28Sjmcp uint_t plw_bin_split_prev; /* last bin we tried to split */ 267d94ffb28Sjmcp uint_t plw_do_split; /* set if OK to split */ 268d94ffb28Sjmcp uint_t plw_split_next; /* next bin to split */ 269d94ffb28Sjmcp uint_t plw_ceq_dif; /* number of different color groups */ 270d94ffb28Sjmcp /* to check */ 271d94ffb28Sjmcp uint_t plw_ceq_mask[MMU_PAGE_SIZES + 1]; /* color equiv mask */ 272d94ffb28Sjmcp uint_t plw_bins[MMU_PAGE_SIZES + 1]; /* num of bins */ 273d94ffb28Sjmcp } page_list_walker_t; 274d94ffb28Sjmcp 2755d07b933Sdp void page_list_walk_init(uchar_t szc, uint_t flags, uint_t bin, 2765d07b933Sdp int can_split, int use_ceq, page_list_walker_t *plw); 2775d07b933Sdp 2785d07b933Sdp uint_t page_list_walk_next_bin(uchar_t szc, uint_t bin, 2795d07b933Sdp page_list_walker_t *plw); 2805d07b933Sdp 281affbd3ccSkchow extern struct cpu cpus[]; 282affbd3ccSkchow #define CPU0 cpus 283affbd3ccSkchow 284843e1988Sjohnlev extern int mtype_init(vnode_t *, caddr_t, uint_t *, size_t); 285843e1988Sjohnlev #define MTYPE_INIT(mtype, vp, vaddr, flags, pgsz) \ 286843e1988Sjohnlev (mtype = mtype_init(vp, vaddr, &(flags), pgsz)) 2877c478bd9Sstevel@tonic-gate 2887c478bd9Sstevel@tonic-gate /* 2897c478bd9Sstevel@tonic-gate * macros to loop through the mtype range (page_get_mnode_{free,cache,any}list, 2907c478bd9Sstevel@tonic-gate * and page_get_contig_pages) 2917c478bd9Sstevel@tonic-gate * 2927c478bd9Sstevel@tonic-gate * MTYPE_START sets the initial mtype. -1 if the mtype range specified does 2937c478bd9Sstevel@tonic-gate * not contain mnode. 2947c478bd9Sstevel@tonic-gate * 2957c478bd9Sstevel@tonic-gate * MTYPE_NEXT sets the next mtype. -1 if there are no more valid 2967c478bd9Sstevel@tonic-gate * mtype in the range. 2977c478bd9Sstevel@tonic-gate */ 2987c478bd9Sstevel@tonic-gate 2997c478bd9Sstevel@tonic-gate #define MTYPE_START(mnode, mtype, flags) \ 3007c478bd9Sstevel@tonic-gate (mtype = mtype_func(mnode, mtype, flags)) 3017c478bd9Sstevel@tonic-gate 302affbd3ccSkchow #define MTYPE_NEXT(mnode, mtype, flags) { \ 303affbd3ccSkchow if (flags & PGI_MT_RANGE) { \ 304affbd3ccSkchow mtype = mtype_func(mnode, mtype, flags | PGI_MT_NEXT); \ 305affbd3ccSkchow } else { \ 306affbd3ccSkchow mtype = -1; \ 307affbd3ccSkchow } \ 308affbd3ccSkchow } 3097c478bd9Sstevel@tonic-gate 310d3b5f563SJohn Levon extern int mtype_pgr_init(int *, page_t *, pgcnt_t); 311d3b5f563SJohn Levon #define MTYPE_PGR_INIT(mtype, flags, pp, pgcnt) \ 312d3b5f563SJohn Levon (mtype = mtype_pgr_init(&flags, pp, pgcnt)) 3137c478bd9Sstevel@tonic-gate 314affbd3ccSkchow #define MNODE_PGCNT(mnode) mnode_pgcnt(mnode) 315affbd3ccSkchow 316843e1988Sjohnlev extern void mnodetype_2_pfn(int, int, pfn_t *, pfn_t *); 3177c478bd9Sstevel@tonic-gate #define MNODETYPE_2_PFN(mnode, mtype, pfnlo, pfnhi) \ 318843e1988Sjohnlev mnodetype_2_pfn(mnode, mtype, &pfnlo, &pfnhi) 3197c478bd9Sstevel@tonic-gate 320d94ffb28Sjmcp #define PC_BIN_MUTEX(mnode, bin, flags) ((flags & PG_FREE_LIST) ? \ 321d94ffb28Sjmcp &fpc_mutex[(bin) & (NPC_MUTEX - 1)][mnode] : \ 3227c478bd9Sstevel@tonic-gate &cpc_mutex[(bin) & (NPC_MUTEX - 1)][mnode]) 3237c478bd9Sstevel@tonic-gate 3247c478bd9Sstevel@tonic-gate #define FPC_MUTEX(mnode, i) (&fpc_mutex[i][mnode]) 3257c478bd9Sstevel@tonic-gate #define CPC_MUTEX(mnode, i) (&cpc_mutex[i][mnode]) 3267c478bd9Sstevel@tonic-gate 3277c478bd9Sstevel@tonic-gate #ifdef DEBUG 3287c478bd9Sstevel@tonic-gate #define CHK_LPG(pp, szc) chk_lpg(pp, szc) 3297c478bd9Sstevel@tonic-gate extern void chk_lpg(page_t *, uchar_t); 3307c478bd9Sstevel@tonic-gate #else 3317c478bd9Sstevel@tonic-gate #define CHK_LPG(pp, szc) 3327c478bd9Sstevel@tonic-gate #endif 3337c478bd9Sstevel@tonic-gate 3347c478bd9Sstevel@tonic-gate #define FULL_REGION_CNT(rg_szc) \ 3357c478bd9Sstevel@tonic-gate (LEVEL_SIZE(rg_szc) >> LEVEL_SHIFT(rg_szc - 1)) 3367c478bd9Sstevel@tonic-gate 3377c478bd9Sstevel@tonic-gate /* Return the leader for this mapping size */ 3387c478bd9Sstevel@tonic-gate #define PP_GROUPLEADER(pp, szc) \ 3397c478bd9Sstevel@tonic-gate (&(pp)[-(int)((pp)->p_pagenum & (SZCPAGES(szc)-1))]) 3407c478bd9Sstevel@tonic-gate 3417c478bd9Sstevel@tonic-gate /* Return the root page for this page based on p_szc */ 3427c478bd9Sstevel@tonic-gate #define PP_PAGEROOT(pp) ((pp)->p_szc == 0 ? (pp) : \ 3437c478bd9Sstevel@tonic-gate PP_GROUPLEADER((pp), (pp)->p_szc)) 3447c478bd9Sstevel@tonic-gate 3457c478bd9Sstevel@tonic-gate /* 3467c478bd9Sstevel@tonic-gate * The counter base must be per page_counter element to prevent 3477c478bd9Sstevel@tonic-gate * races when re-indexing, and the base page size element should 3487c478bd9Sstevel@tonic-gate * be aligned on a boundary of the given region size. 3497c478bd9Sstevel@tonic-gate * 3507c478bd9Sstevel@tonic-gate * We also round up the number of pages spanned by the counters 3517c478bd9Sstevel@tonic-gate * for a given region to PC_BASE_ALIGN in certain situations to simplify 3527c478bd9Sstevel@tonic-gate * the coding for some non-performance critical routines. 3537c478bd9Sstevel@tonic-gate */ 3547c478bd9Sstevel@tonic-gate 3557c478bd9Sstevel@tonic-gate #define PC_BASE_ALIGN ((pfn_t)1 << PAGE_BSZS_SHIFT(MMU_PAGE_SIZES-1)) 3567c478bd9Sstevel@tonic-gate #define PC_BASE_ALIGN_MASK (PC_BASE_ALIGN - 1) 3577c478bd9Sstevel@tonic-gate 3587c478bd9Sstevel@tonic-gate /* 3597c478bd9Sstevel@tonic-gate * cpu/mmu-dependent vm variables 3607c478bd9Sstevel@tonic-gate */ 3617c478bd9Sstevel@tonic-gate extern uint_t mmu_page_sizes; 3627c478bd9Sstevel@tonic-gate extern uint_t mmu_exported_page_sizes; 36302bc52beSkchow /* 36402bc52beSkchow * page sizes that legacy applications can see via getpagesizes(3c). 36502bc52beSkchow * Used to prevent legacy applications from inadvertantly using the 36602bc52beSkchow * 'new' large pagesizes (1g and above). 36702bc52beSkchow */ 36802bc52beSkchow extern uint_t mmu_legacy_page_sizes; 3697c478bd9Sstevel@tonic-gate 3707c478bd9Sstevel@tonic-gate /* For x86, userszc is the same as the kernel's szc */ 3717c478bd9Sstevel@tonic-gate #define USERSZC_2_SZC(userszc) (userszc) 3727c478bd9Sstevel@tonic-gate #define SZC_2_USERSZC(szc) (szc) 3737c478bd9Sstevel@tonic-gate 3747c478bd9Sstevel@tonic-gate /* 3757c478bd9Sstevel@tonic-gate * for hw_page_map_t, sized to hold the ratio of large page to base 3767c478bd9Sstevel@tonic-gate * pagesize (1024 max) 3777c478bd9Sstevel@tonic-gate */ 3787c478bd9Sstevel@tonic-gate typedef short hpmctr_t; 3797c478bd9Sstevel@tonic-gate 3807c478bd9Sstevel@tonic-gate /* 3817c478bd9Sstevel@tonic-gate * get the setsize of the current cpu - assume homogenous for x86 3827c478bd9Sstevel@tonic-gate */ 3837c478bd9Sstevel@tonic-gate extern int l2cache_sz, l2cache_linesz, l2cache_assoc; 3847c478bd9Sstevel@tonic-gate 3857c478bd9Sstevel@tonic-gate #define L2CACHE_ALIGN l2cache_linesz 386affbd3ccSkchow #define L2CACHE_ALIGN_MAX 64 3877c478bd9Sstevel@tonic-gate #define CPUSETSIZE() \ 3887c478bd9Sstevel@tonic-gate (l2cache_assoc ? (l2cache_sz / l2cache_assoc) : MMU_PAGESIZE) 3897c478bd9Sstevel@tonic-gate 3907c478bd9Sstevel@tonic-gate /* 3917c478bd9Sstevel@tonic-gate * Return the log2(pagesize(szc) / MMU_PAGESIZE) --- or the shift count 3927c478bd9Sstevel@tonic-gate * for the number of base pages in this pagesize 3937c478bd9Sstevel@tonic-gate */ 3947c478bd9Sstevel@tonic-gate #define PAGE_BSZS_SHIFT(szc) (LEVEL_SHIFT(szc) - MMU_PAGESHIFT) 3957c478bd9Sstevel@tonic-gate 3967c478bd9Sstevel@tonic-gate /* 3977c478bd9Sstevel@tonic-gate * Internal PG_ flags. 3987c478bd9Sstevel@tonic-gate */ 3997c478bd9Sstevel@tonic-gate #define PGI_RELOCONLY 0x010000 /* opposite of PG_NORELOC */ 4007c478bd9Sstevel@tonic-gate #define PGI_NOCAGE 0x020000 /* cage is disabled */ 4017c478bd9Sstevel@tonic-gate #define PGI_PGCPHIPRI 0x040000 /* page_get_contig_page pri alloc */ 4027c478bd9Sstevel@tonic-gate #define PGI_PGCPSZC0 0x080000 /* relocate base pagesize page */ 4037c478bd9Sstevel@tonic-gate 4047c478bd9Sstevel@tonic-gate /* 4057c478bd9Sstevel@tonic-gate * PGI range flags - should not overlap PGI flags 4067c478bd9Sstevel@tonic-gate */ 4077c478bd9Sstevel@tonic-gate #define PGI_MT_RANGE0 0x1000000 /* mtype range to 0 */ 408843e1988Sjohnlev #define PGI_MT_RANGE16M 0x2000000 /* mtype range to 16m */ 40907ad560dSkchow #define PGI_MT_RANGE4G 0x4000000 /* mtype range to 4g */ 41007ad560dSkchow #define PGI_MT_NEXT 0x8000000 /* get next mtype */ 41107ad560dSkchow #define PGI_MT_RANGE (PGI_MT_RANGE0 | PGI_MT_RANGE16M | PGI_MT_RANGE4G) 4127c478bd9Sstevel@tonic-gate 413843e1988Sjohnlev 414ec25b48fSsusans /* 415ec25b48fSsusans * Maximum and default values for user heap, stack, private and shared 416ec25b48fSsusans * anonymous memory, and user text and initialized data. 417ec25b48fSsusans * Used by map_pgsz*() routines. 418ec25b48fSsusans */ 419ec25b48fSsusans extern size_t max_uheap_lpsize; 420ec25b48fSsusans extern size_t default_uheap_lpsize; 421ec25b48fSsusans extern size_t max_ustack_lpsize; 422ec25b48fSsusans extern size_t default_ustack_lpsize; 423ec25b48fSsusans extern size_t max_privmap_lpsize; 424ec25b48fSsusans extern size_t max_uidata_lpsize; 425ec25b48fSsusans extern size_t max_utext_lpsize; 426ec25b48fSsusans extern size_t max_shm_lpsize; 427ec25b48fSsusans extern size_t mcntl0_lpsize; 428ec25b48fSsusans 429ec25b48fSsusans /* 430ec25b48fSsusans * Sanity control. Don't use large pages regardless of user 431ec25b48fSsusans * settings if there's less than priv or shm_lpg_min_physmem memory installed. 432ec25b48fSsusans * The units for this variable are 8K pages. 433ec25b48fSsusans */ 434ec25b48fSsusans extern pgcnt_t privm_lpg_min_physmem; 435ec25b48fSsusans extern pgcnt_t shm_lpg_min_physmem; 436ec25b48fSsusans 4377c478bd9Sstevel@tonic-gate /* 4387c478bd9Sstevel@tonic-gate * hash as and addr to get a bin. 4397c478bd9Sstevel@tonic-gate */ 4407c478bd9Sstevel@tonic-gate 441d94ffb28Sjmcp #define AS_2_BIN(as, seg, vp, addr, bin, szc) \ 4425d07b933Sdp bin = (((((uintptr_t)(addr) >> PAGESHIFT) + ((uintptr_t)(as) >> 4)) \ 4435d07b933Sdp & page_colors_mask) >> \ 4445d07b933Sdp (hw_page_array[szc].hp_shift - hw_page_array[0].hp_shift)) 4457c478bd9Sstevel@tonic-gate 446affbd3ccSkchow /* 447affbd3ccSkchow * cpu private vm data - accessed thru CPU->cpu_vm_data 448affbd3ccSkchow * vc_pnum_memseg: tracks last memseg visited in page_numtopp_nolock() 449affbd3ccSkchow * vc_pnext_memseg: tracks last memseg visited in page_nextn() 450affbd3ccSkchow * vc_kmptr: orignal unaligned kmem pointer for this vm_cpu_data_t 4516061ce8aSkchow * vc_kmsize: orignal kmem size for this vm_cpu_data_t 452affbd3ccSkchow */ 453affbd3ccSkchow 454affbd3ccSkchow typedef struct { 455affbd3ccSkchow struct memseg *vc_pnum_memseg; 456affbd3ccSkchow struct memseg *vc_pnext_memseg; 457affbd3ccSkchow void *vc_kmptr; 4586061ce8aSkchow size_t vc_kmsize; 459affbd3ccSkchow } vm_cpu_data_t; 460affbd3ccSkchow 461affbd3ccSkchow /* allocation size to ensure vm_cpu_data_t resides in its own cache line */ 462affbd3ccSkchow #define VM_CPU_DATA_PADSIZE \ 463affbd3ccSkchow (P2ROUNDUP(sizeof (vm_cpu_data_t), L2CACHE_ALIGN_MAX)) 464affbd3ccSkchow 4657c478bd9Sstevel@tonic-gate /* 4667c478bd9Sstevel@tonic-gate * When a bin is empty, and we can't satisfy a color request correctly, 4677c478bd9Sstevel@tonic-gate * we scan. If we assume that the programs have reasonable spatial 4687c478bd9Sstevel@tonic-gate * behavior, then it will not be a good idea to use the adjacent color. 4697c478bd9Sstevel@tonic-gate * Using the adjacent color would result in virtually adjacent addresses 4707c478bd9Sstevel@tonic-gate * mapping into the same spot in the cache. So, if we stumble across 4717c478bd9Sstevel@tonic-gate * an empty bin, skip a bunch before looking. After the first skip, 4727c478bd9Sstevel@tonic-gate * then just look one bin at a time so we don't miss our cache on 4737c478bd9Sstevel@tonic-gate * every look. Be sure to check every bin. Page_create() will panic 4747c478bd9Sstevel@tonic-gate * if we miss a page. 4757c478bd9Sstevel@tonic-gate * 4767c478bd9Sstevel@tonic-gate * This also explains the `<=' in the for loops in both page_get_freelist() 4777c478bd9Sstevel@tonic-gate * and page_get_cachelist(). Since we checked the target bin, skipped 4787c478bd9Sstevel@tonic-gate * a bunch, then continued one a time, we wind up checking the target bin 4797c478bd9Sstevel@tonic-gate * twice to make sure we get all of them bins. 4807c478bd9Sstevel@tonic-gate */ 4817c478bd9Sstevel@tonic-gate #define BIN_STEP 19 4827c478bd9Sstevel@tonic-gate 4837c478bd9Sstevel@tonic-gate #ifdef VM_STATS 4847c478bd9Sstevel@tonic-gate struct vmm_vmstats_str { 485d94ffb28Sjmcp ulong_t pgf_alloc[MMU_PAGE_SIZES]; /* page_get_freelist */ 486d94ffb28Sjmcp ulong_t pgf_allocok[MMU_PAGE_SIZES]; 487d94ffb28Sjmcp ulong_t pgf_allocokrem[MMU_PAGE_SIZES]; 488d94ffb28Sjmcp ulong_t pgf_allocfailed[MMU_PAGE_SIZES]; 4897c478bd9Sstevel@tonic-gate ulong_t pgf_allocdeferred; 490d94ffb28Sjmcp ulong_t pgf_allocretry[MMU_PAGE_SIZES]; 491affbd3ccSkchow ulong_t pgc_alloc; /* page_get_cachelist */ 4927c478bd9Sstevel@tonic-gate ulong_t pgc_allocok; 4937c478bd9Sstevel@tonic-gate ulong_t pgc_allocokrem; 4947c478bd9Sstevel@tonic-gate ulong_t pgc_allocokdeferred; 4957c478bd9Sstevel@tonic-gate ulong_t pgc_allocfailed; 496affbd3ccSkchow ulong_t pgcp_alloc[MMU_PAGE_SIZES]; /* page_get_contig_pages */ 4977c478bd9Sstevel@tonic-gate ulong_t pgcp_allocfailed[MMU_PAGE_SIZES]; 4987c478bd9Sstevel@tonic-gate ulong_t pgcp_allocempty[MMU_PAGE_SIZES]; 4997c478bd9Sstevel@tonic-gate ulong_t pgcp_allocok[MMU_PAGE_SIZES]; 500affbd3ccSkchow ulong_t ptcp[MMU_PAGE_SIZES]; /* page_trylock_contig_pages */ 5017c478bd9Sstevel@tonic-gate ulong_t ptcpfreethresh[MMU_PAGE_SIZES]; 5027c478bd9Sstevel@tonic-gate ulong_t ptcpfailexcl[MMU_PAGE_SIZES]; 5037c478bd9Sstevel@tonic-gate ulong_t ptcpfailszc[MMU_PAGE_SIZES]; 5047c478bd9Sstevel@tonic-gate ulong_t ptcpfailcage[MMU_PAGE_SIZES]; 5057c478bd9Sstevel@tonic-gate ulong_t ptcpok[MMU_PAGE_SIZES]; 506affbd3ccSkchow ulong_t pgmf_alloc[MMU_PAGE_SIZES]; /* page_get_mnode_freelist */ 5077c478bd9Sstevel@tonic-gate ulong_t pgmf_allocfailed[MMU_PAGE_SIZES]; 5087c478bd9Sstevel@tonic-gate ulong_t pgmf_allocempty[MMU_PAGE_SIZES]; 5097c478bd9Sstevel@tonic-gate ulong_t pgmf_allocok[MMU_PAGE_SIZES]; 510affbd3ccSkchow ulong_t pgmc_alloc; /* page_get_mnode_cachelist */ 5117c478bd9Sstevel@tonic-gate ulong_t pgmc_allocfailed; 5127c478bd9Sstevel@tonic-gate ulong_t pgmc_allocempty; 5137c478bd9Sstevel@tonic-gate ulong_t pgmc_allocok; 514affbd3ccSkchow ulong_t pladd_free[MMU_PAGE_SIZES]; /* page_list_add/sub */ 515affbd3ccSkchow ulong_t plsub_free[MMU_PAGE_SIZES]; 516affbd3ccSkchow ulong_t pladd_cache; 517affbd3ccSkchow ulong_t plsub_cache; 518affbd3ccSkchow ulong_t plsubpages_szcbig; 519affbd3ccSkchow ulong_t plsubpages_szc0; 5205d07b933Sdp ulong_t pfs_req[MMU_PAGE_SIZES]; /* page_freelist_split */ 5215d07b933Sdp ulong_t pfs_demote[MMU_PAGE_SIZES]; 5225d07b933Sdp ulong_t pfc_coalok[MMU_PAGE_SIZES][MAX_MNODE_MRANGES]; 523affbd3ccSkchow ulong_t ppr_reloc[MMU_PAGE_SIZES]; /* page_relocate */ 5247c478bd9Sstevel@tonic-gate ulong_t ppr_relocnoroot[MMU_PAGE_SIZES]; 5257c478bd9Sstevel@tonic-gate ulong_t ppr_reloc_replnoroot[MMU_PAGE_SIZES]; 5267c478bd9Sstevel@tonic-gate ulong_t ppr_relocnolock[MMU_PAGE_SIZES]; 5277c478bd9Sstevel@tonic-gate ulong_t ppr_relocnomem[MMU_PAGE_SIZES]; 5287c478bd9Sstevel@tonic-gate ulong_t ppr_relocok[MMU_PAGE_SIZES]; 5298b464eb8Smec ulong_t ppr_copyfail; 5305d07b933Sdp /* page coalesce counter */ 5315d07b933Sdp ulong_t page_ctrs_coalesce[MMU_PAGE_SIZES][MAX_MNODE_MRANGES]; 5325d07b933Sdp /* candidates useful */ 5335d07b933Sdp ulong_t page_ctrs_cands_skip[MMU_PAGE_SIZES][MAX_MNODE_MRANGES]; 5345d07b933Sdp /* ctrs changed after locking */ 5355d07b933Sdp ulong_t page_ctrs_changed[MMU_PAGE_SIZES][MAX_MNODE_MRANGES]; 5365d07b933Sdp /* page_freelist_coalesce failed */ 5375d07b933Sdp ulong_t page_ctrs_failed[MMU_PAGE_SIZES][MAX_MNODE_MRANGES]; 5387c478bd9Sstevel@tonic-gate ulong_t page_ctrs_coalesce_all; /* page coalesce all counter */ 5397c478bd9Sstevel@tonic-gate ulong_t page_ctrs_cands_skip_all; /* candidates useful for all func */ 5407c478bd9Sstevel@tonic-gate ulong_t restrict4gcnt; 54107ad560dSkchow ulong_t unrestrict16mcnt; /* non-DMA 16m allocs allowed */ 542a985e578Skchow ulong_t pgpanicalloc; /* PG_PANIC allocation */ 54378b03d3aSkchow ulong_t pcf_deny[MMU_PAGE_SIZES]; /* page_chk_freelist */ 54478b03d3aSkchow ulong_t pcf_allow[MMU_PAGE_SIZES]; 5457c478bd9Sstevel@tonic-gate }; 5467c478bd9Sstevel@tonic-gate extern struct vmm_vmstats_str vmm_vmstats; 5477c478bd9Sstevel@tonic-gate #endif /* VM_STATS */ 5487c478bd9Sstevel@tonic-gate 5497c478bd9Sstevel@tonic-gate extern size_t page_ctrs_sz(void); 5507c478bd9Sstevel@tonic-gate extern caddr_t page_ctrs_alloc(caddr_t); 551affbd3ccSkchow extern void page_ctr_sub(int, int, page_t *, int); 5525d07b933Sdp extern page_t *page_freelist_split(uchar_t, 55319397407SSherry Moore uint_t, int, int, pfn_t, pfn_t, page_list_walker_t *); 5545d07b933Sdp extern page_t *page_freelist_coalesce(int, uchar_t, uint_t, uint_t, int, 5555d07b933Sdp pfn_t); 556a3114836SGerry Liu extern void page_freelist_coalesce_all(int); 5577c478bd9Sstevel@tonic-gate extern uint_t page_get_pagecolors(uint_t); 558d2b85481Srscott extern void pfnzero(pfn_t, uint_t, uint_t); 5597c478bd9Sstevel@tonic-gate 5607c478bd9Sstevel@tonic-gate #ifdef __cplusplus 5617c478bd9Sstevel@tonic-gate } 5627c478bd9Sstevel@tonic-gate #endif 5637c478bd9Sstevel@tonic-gate 5647c478bd9Sstevel@tonic-gate #endif /* _VM_DEP_H */ 565