/* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 1986, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2015, Josef 'Jeff' Sipek * Copyright (c) 2015, 2016 by Delphix. All rights reserved. * Copyright 2018 Joyent, Inc. * Copyright 2021 Oxide Computer Company */ /* Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T */ /* All Rights Reserved */ /* * University Copyright- Copyright (c) 1982, 1986, 1988 * The Regents of the University of California * All Rights Reserved * * University Acknowledgment- Portions of this document are derived from * software developed by the University of California, Berkeley, and its * contributors. */ /* * VM - physical page management. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static pgcnt_t max_page_get; /* max page_get request size in pages */ pgcnt_t total_pages = 0; /* total number of pages (used by /proc) */ volatile uint64_t n_throttle = 0; /* * freemem_lock protects all freemem variables: * availrmem. Also this lock protects the globals which track the * availrmem changes for accurate kernel footprint calculation. * See below for an explanation of these * globals. */ kmutex_t freemem_lock; pgcnt_t availrmem; pgcnt_t availrmem_initial; /* * These globals track availrmem changes to get a more accurate * estimate of tke kernel size. Historically pp_kernel is used for * kernel size and is based on availrmem. But availrmem is adjusted for * locked pages in the system not just for kernel locked pages. * These new counters will track the pages locked through segvn and * by explicit user locking. * * pages_locked : How many pages are locked because of user specified * locking through mlock or plock. * * pages_useclaim,pages_claimed : These two variables track the * claim adjustments because of the protection changes on a segvn segment. * * All these globals are protected by the same lock which protects availrmem. */ pgcnt_t pages_locked = 0; pgcnt_t pages_useclaim = 0; pgcnt_t pages_claimed = 0; /* * new_freemem_lock protects freemem, freemem_wait & freemem_cv. */ static kmutex_t new_freemem_lock; static uint_t freemem_wait; /* someone waiting for freemem */ static kcondvar_t freemem_cv; /* * The logical page free list is maintained as two lists, the 'free' * and the 'cache' lists. * The free list contains those pages that should be reused first. * * The implementation of the lists is machine dependent. * page_get_freelist(), page_get_cachelist(), * page_list_sub(), and page_list_add() * form the interface to the machine dependent implementation. * * Pages with p_free set are on the cache list. * Pages with p_free and p_age set are on the free list, * * A page may be locked while on either list. */ /* * free list accounting stuff. * * * Spread out the value for the number of pages on the * page free and page cache lists. If there is just one * value, then it must be under just one lock. * The lock contention and cache traffic are a real bother. * * When we acquire and then drop a single pcf lock * we can start in the middle of the array of pcf structures. * If we acquire more than one pcf lock at a time, we need to * start at the front to avoid deadlocking. * * pcf_count holds the number of pages in each pool. * * pcf_block is set when page_create_get_something() has asked the * PSM page freelist and page cachelist routines without specifying * a color and nothing came back. This is used to block anything * else from moving pages from one list to the other while the * lists are searched again. If a page is freeed while pcf_block is * set, then pcf_reserve is incremented. pcgs_unblock() takes care * of clearning pcf_block, doing the wakeups, etc. */ #define MAX_PCF_FANOUT NCPU static uint_t pcf_fanout = 1; /* Will get changed at boot time */ static uint_t pcf_fanout_mask = 0; struct pcf { kmutex_t pcf_lock; /* protects the structure */ uint_t pcf_count; /* page count */ uint_t pcf_wait; /* number of waiters */ uint_t pcf_block; /* pcgs flag to page_free() */ uint_t pcf_reserve; /* pages freed after pcf_block set */ uint_t pcf_fill[10]; /* to line up on the caches */ }; /* * PCF_INDEX hash needs to be dynamic (every so often the hash changes where * it will hash the cpu to). This is done to prevent a drain condition * from happening. This drain condition will occur when pcf_count decrement * occurs on cpu A and the increment of pcf_count always occurs on cpu B. An * example of this shows up with device interrupts. The dma buffer is allocated * by the cpu requesting the IO thus the pcf_count is decremented based on that. * When the memory is returned by the interrupt thread, the pcf_count will be * incremented based on the cpu servicing the interrupt. */ static struct pcf pcf[MAX_PCF_FANOUT]; #define PCF_INDEX() ((int)(((long)CPU->cpu_seqid) + \ (randtick() >> 24)) & (pcf_fanout_mask)) static int pcf_decrement_bucket(pgcnt_t); static int pcf_decrement_multiple(pgcnt_t *, pgcnt_t, int); kmutex_t pcgs_lock; /* serializes page_create_get_ */ kmutex_t pcgs_cagelock; /* serializes NOSLEEP cage allocs */ kmutex_t pcgs_wait_lock; /* used for delay in pcgs */ static kcondvar_t pcgs_cv; /* cv for delay in pcgs */ #ifdef VM_STATS /* * No locks, but so what, they are only statistics. */ static struct page_tcnt { int pc_free_cache; /* free's into cache list */ int pc_free_dontneed; /* free's with dontneed */ int pc_free_pageout; /* free's from pageout */ int pc_free_free; /* free's into free list */ int pc_free_pages; /* free's into large page free list */ int pc_destroy_pages; /* large page destroy's */ int pc_get_cache; /* get's from cache list */ int pc_get_free; /* get's from free list */ int pc_reclaim; /* reclaim's */ int pc_abortfree; /* abort's of free pages */ int pc_find_hit; /* find's that find page */ int pc_find_miss; /* find's that don't find page */ int pc_destroy_free; /* # of free pages destroyed */ #define PC_HASH_CNT (4*PAGE_HASHAVELEN) int pc_find_hashlen[PC_HASH_CNT+1]; int pc_addclaim_pages; int pc_subclaim_pages; int pc_free_replacement_page[2]; int pc_try_demote_pages[6]; int pc_demote_pages[2]; } pagecnt; uint_t hashin_count; uint_t hashin_not_held; uint_t hashin_already; uint_t hashout_count; uint_t hashout_not_held; uint_t page_create_count; uint_t page_create_not_enough; uint_t page_create_not_enough_again; uint_t page_create_zero; uint_t page_create_hashout; uint_t page_create_page_lock_failed; uint_t page_create_trylock_failed; uint_t page_create_found_one; uint_t page_create_hashin_failed; uint_t page_create_dropped_phm; uint_t page_create_new; uint_t page_create_exists; uint_t page_create_putbacks; uint_t page_create_overshoot; uint_t page_reclaim_zero; uint_t page_reclaim_zero_locked; uint_t page_rename_exists; uint_t page_rename_count; uint_t page_lookup_cnt[20]; uint_t page_lookup_nowait_cnt[10]; uint_t page_find_cnt; uint_t page_exists_cnt; uint_t page_exists_forreal_cnt; uint_t page_lookup_dev_cnt; uint_t get_cachelist_cnt; uint_t page_create_cnt[10]; uint_t alloc_pages[9]; uint_t page_exphcontg[19]; uint_t page_create_large_cnt[10]; #endif static inline page_t * page_hash_search(ulong_t index, vnode_t *vnode, u_offset_t off) { uint_t mylen = 0; page_t *page; for (page = page_hash[index]; page; page = page->p_hash, mylen++) if (page->p_vnode == vnode && page->p_offset == off) break; #ifdef VM_STATS if (page != NULL) pagecnt.pc_find_hit++; else pagecnt.pc_find_miss++; pagecnt.pc_find_hashlen[MIN(mylen, PC_HASH_CNT)]++; #endif return (page); } #ifdef DEBUG #define MEMSEG_SEARCH_STATS #endif #ifdef MEMSEG_SEARCH_STATS struct memseg_stats { uint_t nsearch; uint_t nlastwon; uint_t nhashwon; uint_t nnotfound; } memseg_stats; #define MEMSEG_STAT_INCR(v) \ atomic_inc_32(&memseg_stats.v) #else #define MEMSEG_STAT_INCR(x) #endif struct memseg *memsegs; /* list of memory segments */ /* * /etc/system tunable to control large page allocation hueristic. * * Setting to LPAP_LOCAL will heavily prefer the local lgroup over remote lgroup * for large page allocation requests. If a large page is not readily * avaliable on the local freelists we will go through additional effort * to create a large page, potentially moving smaller pages around to coalesce * larger pages in the local lgroup. * Default value of LPAP_DEFAULT will go to remote freelists if large pages * are not readily available in the local lgroup. */ enum lpap { LPAP_DEFAULT, /* default large page allocation policy */ LPAP_LOCAL /* local large page allocation policy */ }; enum lpap lpg_alloc_prefer = LPAP_DEFAULT; static void page_init_mem_config(void); static int page_do_hashin(page_t *, vnode_t *, u_offset_t); static void page_do_hashout(page_t *); static void page_capture_init(); int page_capture_take_action(page_t *, uint_t, void *); static void page_demote_vp_pages(page_t *); void pcf_init(void) { if (boot_ncpus != -1) { pcf_fanout = boot_ncpus; } else { pcf_fanout = max_ncpus; } #ifdef sun4v /* * Force at least 4 buckets if possible for sun4v. */ pcf_fanout = MAX(pcf_fanout, 4); #endif /* sun4v */ /* * Round up to the nearest power of 2. */ pcf_fanout = MIN(pcf_fanout, MAX_PCF_FANOUT); if (!ISP2(pcf_fanout)) { pcf_fanout = 1 << highbit(pcf_fanout); if (pcf_fanout > MAX_PCF_FANOUT) { pcf_fanout = 1 << (highbit(MAX_PCF_FANOUT) - 1); } } pcf_fanout_mask = pcf_fanout - 1; } /* * vm subsystem related initialization */ void vm_init(void) { boolean_t callb_vm_cpr(void *, int); (void) callb_add(callb_vm_cpr, 0, CB_CL_CPR_VM, "vm"); page_init_mem_config(); page_retire_init(); vm_usage_init(); page_capture_init(); } /* * This function is called at startup and when memory is added or deleted. */ void init_pages_pp_maximum() { static pgcnt_t p_min; static pgcnt_t pages_pp_maximum_startup; static pgcnt_t avrmem_delta; static int init_done; static int user_set; /* true if set in /etc/system */ if (init_done == 0) { /* If the user specified a value, save it */ if (pages_pp_maximum != 0) { user_set = 1; pages_pp_maximum_startup = pages_pp_maximum; } /* * Setting of pages_pp_maximum is based first time * on the value of availrmem just after the start-up * allocations. To preserve this relationship at run * time, use a delta from availrmem_initial. */ ASSERT(availrmem_initial >= availrmem); avrmem_delta = availrmem_initial - availrmem; /* The allowable floor of pages_pp_maximum */ p_min = tune.t_minarmem + 100; /* Make sure we don't come through here again. */ init_done = 1; } /* * Determine pages_pp_maximum, the number of currently available * pages (availrmem) that can't be `locked'. If not set by * the user, we set it to 4% of the currently available memory * plus 4MB. * But we also insist that it be greater than tune.t_minarmem; * otherwise a process could lock down a lot of memory, get swapped * out, and never have enough to get swapped back in. */ if (user_set) pages_pp_maximum = pages_pp_maximum_startup; else pages_pp_maximum = ((availrmem_initial - avrmem_delta) / 25) + btop(4 * 1024 * 1024); if (pages_pp_maximum <= p_min) { pages_pp_maximum = p_min; } } /* * In the past, we limited the maximum pages that could be gotten to essentially * 1/2 of the total pages on the system. However, this is too conservative for * some cases. For example, if we want to host a large virtual machine which * needs to use a significant portion of the system's memory. In practice, * allowing more than 1/2 of the total pages is fine, but becomes problematic * as we approach or exceed 75% of the pages on the system. Thus, we limit the * maximum to 23/32 of the total pages, which is ~72%. */ void set_max_page_get(pgcnt_t target_total_pages) { max_page_get = (target_total_pages >> 5) * 23; ASSERT3U(max_page_get, >, 0); } pgcnt_t get_max_page_get() { return (max_page_get); } static pgcnt_t pending_delete; /*ARGSUSED*/ static void page_mem_config_post_add( void *arg, pgcnt_t delta_pages) { set_max_page_get(total_pages - pending_delete); init_pages_pp_maximum(); } /*ARGSUSED*/ static int page_mem_config_pre_del( void *arg, pgcnt_t delta_pages) { pgcnt_t nv; nv = atomic_add_long_nv(&pending_delete, (spgcnt_t)delta_pages); set_max_page_get(total_pages - nv); return (0); } /*ARGSUSED*/ static void page_mem_config_post_del( void *arg, pgcnt_t delta_pages, int cancelled) { pgcnt_t nv; nv = atomic_add_long_nv(&pending_delete, -(spgcnt_t)delta_pages); set_max_page_get(total_pages - nv); if (!cancelled) init_pages_pp_maximum(); } static kphysm_setup_vector_t page_mem_config_vec = { KPHYSM_SETUP_VECTOR_VERSION, page_mem_config_post_add, page_mem_config_pre_del, page_mem_config_post_del, }; static void page_init_mem_config(void) { int ret; ret = kphysm_setup_func_register(&page_mem_config_vec, (void *)NULL); ASSERT(ret == 0); } /* * Evenly spread out the PCF counters for large free pages */ static void page_free_large_ctr(pgcnt_t npages) { static struct pcf *p = pcf; pgcnt_t lump; freemem += npages; lump = roundup(npages, pcf_fanout) / pcf_fanout; while (npages > 0) { ASSERT(!p->pcf_block); if (lump < npages) { p->pcf_count += (uint_t)lump; npages -= lump; } else { p->pcf_count += (uint_t)npages; npages = 0; } ASSERT(!p->pcf_wait); if (++p > &pcf[pcf_fanout - 1]) p = pcf; } ASSERT(npages == 0); } /* * Add a physical chunk of memory to the system free lists during startup. * Platform specific startup() allocates the memory for the page structs. * * num - number of page structures * base - page number (pfn) to be associated with the first page. * * Since we are doing this during startup (ie. single threaded), we will * use shortcut routines to avoid any locking overhead while putting all * these pages on the freelists. * * NOTE: Any changes performed to page_free(), must also be performed to * add_physmem() since this is how we initialize all page_t's at * boot time. */ void add_physmem( page_t *pp, pgcnt_t num, pfn_t pnum) { page_t *root = NULL; uint_t szc = page_num_pagesizes() - 1; pgcnt_t large = page_get_pagecnt(szc); pgcnt_t cnt = 0; TRACE_2(TR_FAC_VM, TR_PAGE_INIT, "add_physmem:pp %p num %lu", pp, num); /* * Arbitrarily limit the max page_get request * to 1/2 of the page structs we have. */ total_pages += num; set_max_page_get(total_pages); PLCNT_MODIFY_MAX(pnum, (long)num); /* * The physical space for the pages array * representing ram pages has already been * allocated. Here we initialize each lock * in the page structure, and put each on * the free list */ for (; num; pp++, pnum++, num--) { /* * this needs to fill in the page number * and do any other arch specific initialization */ add_physmem_cb(pp, pnum); pp->p_lckcnt = 0; pp->p_cowcnt = 0; pp->p_slckcnt = 0; /* * Initialize the page lock as unlocked, since nobody * can see or access this page yet. */ pp->p_selock = 0; /* * Initialize IO lock */ page_iolock_init(pp); /* * initialize other fields in the page_t */ PP_SETFREE(pp); page_clr_all_props(pp); PP_SETAGED(pp); pp->p_offset = (u_offset_t)-1; pp->p_next = pp; pp->p_prev = pp; /* * Simple case: System doesn't support large pages. */ if (szc == 0) { pp->p_szc = 0; page_free_at_startup(pp); continue; } /* * Handle unaligned pages, we collect them up onto * the root page until we have a full large page. */ if (!IS_P2ALIGNED(pnum, large)) { /* * If not in a large page, * just free as small page. */ if (root == NULL) { pp->p_szc = 0; page_free_at_startup(pp); continue; } /* * Link a constituent page into the large page. */ pp->p_szc = szc; page_list_concat(&root, &pp); /* * When large page is fully formed, free it. */ if (++cnt == large) { page_free_large_ctr(cnt); page_list_add_pages(root, PG_LIST_ISINIT); root = NULL; cnt = 0; } continue; } /* * At this point we have a page number which * is aligned. We assert that we aren't already * in a different large page. */ ASSERT(IS_P2ALIGNED(pnum, large)); ASSERT(root == NULL && cnt == 0); /* * If insufficient number of pages left to form * a large page, just free the small page. */ if (num < large) { pp->p_szc = 0; page_free_at_startup(pp); continue; } /* * Otherwise start a new large page. */ pp->p_szc = szc; cnt++; root = pp; } ASSERT(root == NULL && cnt == 0); } /* * Find a page representing the specified [vp, offset]. * If we find the page but it is intransit coming in, * it will have an "exclusive" lock and we wait for * the i/o to complete. A page found on the free list * is always reclaimed and then locked. On success, the page * is locked, its data is valid and it isn't on the free * list, while a NULL is returned if the page doesn't exist. */ page_t * page_lookup(vnode_t *vp, u_offset_t off, se_t se) { return (page_lookup_create(vp, off, se, NULL, NULL, 0)); } /* * Find a page representing the specified [vp, offset]. * We either return the one we found or, if passed in, * create one with identity of [vp, offset] of the * pre-allocated page. If we find existing page but it is * intransit coming in, it will have an "exclusive" lock * and we wait for the i/o to complete. A page found on * the free list is always reclaimed and then locked. * On success, the page is locked, its data is valid and * it isn't on the free list, while a NULL is returned * if the page doesn't exist and newpp is NULL; */ page_t * page_lookup_create( vnode_t *vp, u_offset_t off, se_t se, page_t *newpp, spgcnt_t *nrelocp, int flags) { page_t *pp; kmutex_t *phm; ulong_t index; uint_t hash_locked; uint_t es; ASSERT(MUTEX_NOT_HELD(page_vnode_mutex(vp))); VM_STAT_ADD(page_lookup_cnt[0]); ASSERT(newpp ? PAGE_EXCL(newpp) : 1); /* * Acquire the appropriate page hash lock since * we have to search the hash list. Pages that * hash to this list can't change identity while * this lock is held. */ hash_locked = 0; index = PAGE_HASH_FUNC(vp, off); phm = NULL; top: pp = page_hash_search(index, vp, off); if (pp != NULL) { VM_STAT_ADD(page_lookup_cnt[1]); es = (newpp != NULL) ? 1 : 0; es |= flags; if (!hash_locked) { VM_STAT_ADD(page_lookup_cnt[2]); if (!page_try_reclaim_lock(pp, se, es)) { /* * On a miss, acquire the phm. Then * next time, page_lock() will be called, * causing a wait if the page is busy. * just looping with page_trylock() would * get pretty boring. */ VM_STAT_ADD(page_lookup_cnt[3]); phm = PAGE_HASH_MUTEX(index); mutex_enter(phm); hash_locked = 1; goto top; } } else { VM_STAT_ADD(page_lookup_cnt[4]); if (!page_lock_es(pp, se, phm, P_RECLAIM, es)) { VM_STAT_ADD(page_lookup_cnt[5]); goto top; } } /* * Since `pp' is locked it can not change identity now. * Reconfirm we locked the correct page. * * Both the p_vnode and p_offset *must* be cast volatile * to force a reload of their values: The page_hash_search * function will have stuffed p_vnode and p_offset into * registers before calling page_trylock(); another thread, * actually holding the hash lock, could have changed the * page's identity in memory, but our registers would not * be changed, fooling the reconfirmation. If the hash * lock was held during the search, the casting would * not be needed. */ VM_STAT_ADD(page_lookup_cnt[6]); if (((volatile struct vnode *)(pp->p_vnode) != vp) || ((volatile u_offset_t)(pp->p_offset) != off)) { VM_STAT_ADD(page_lookup_cnt[7]); if (hash_locked) { panic("page_lookup_create: lost page %p", (void *)pp); /*NOTREACHED*/ } page_unlock(pp); phm = PAGE_HASH_MUTEX(index); mutex_enter(phm); hash_locked = 1; goto top; } /* * If page_trylock() was called, then pp may still be on * the cachelist (can't be on the free list, it would not * have been found in the search). If it is on the * cachelist it must be pulled now. To pull the page from * the cachelist, it must be exclusively locked. * * The other big difference between page_trylock() and * page_lock(), is that page_lock() will pull the * page from whatever free list (the cache list in this * case) the page is on. If page_trylock() was used * above, then we have to do the reclaim ourselves. */ if ((!hash_locked) && (PP_ISFREE(pp))) { ASSERT(PP_ISAGED(pp) == 0); VM_STAT_ADD(page_lookup_cnt[8]); /* * page_relcaim will insure that we * have this page exclusively */ if (!page_reclaim(pp, NULL)) { /* * Page_reclaim dropped whatever lock * we held. */ VM_STAT_ADD(page_lookup_cnt[9]); phm = PAGE_HASH_MUTEX(index); mutex_enter(phm); hash_locked = 1; goto top; } else if (se == SE_SHARED && newpp == NULL) { VM_STAT_ADD(page_lookup_cnt[10]); page_downgrade(pp); } } if (hash_locked) { mutex_exit(phm); } if (newpp != NULL && pp->p_szc < newpp->p_szc && PAGE_EXCL(pp) && nrelocp != NULL) { ASSERT(nrelocp != NULL); (void) page_relocate(&pp, &newpp, 1, 1, nrelocp, NULL); if (*nrelocp > 0) { VM_STAT_COND_ADD(*nrelocp == 1, page_lookup_cnt[11]); VM_STAT_COND_ADD(*nrelocp > 1, page_lookup_cnt[12]); pp = newpp; se = SE_EXCL; } else { if (se == SE_SHARED) { page_downgrade(pp); } VM_STAT_ADD(page_lookup_cnt[13]); } } else if (newpp != NULL && nrelocp != NULL) { if (PAGE_EXCL(pp) && se == SE_SHARED) { page_downgrade(pp); } VM_STAT_COND_ADD(pp->p_szc < newpp->p_szc, page_lookup_cnt[14]); VM_STAT_COND_ADD(pp->p_szc == newpp->p_szc, page_lookup_cnt[15]); VM_STAT_COND_ADD(pp->p_szc > newpp->p_szc, page_lookup_cnt[16]); } else if (newpp != NULL && PAGE_EXCL(pp)) { se = SE_EXCL; } } else if (!hash_locked) { VM_STAT_ADD(page_lookup_cnt[17]); phm = PAGE_HASH_MUTEX(index); mutex_enter(phm); hash_locked = 1; goto top; } else if (newpp != NULL) { /* * If we have a preallocated page then * insert it now and basically behave like * page_create. */ VM_STAT_ADD(page_lookup_cnt[18]); /* * Since we hold the page hash mutex and * just searched for this page, page_hashin * had better not fail. If it does, that * means some thread did not follow the * page hash mutex rules. Panic now and * get it over with. As usual, go down * holding all the locks. */ ASSERT(MUTEX_HELD(phm)); if (!page_hashin(newpp, vp, off, phm)) { ASSERT(MUTEX_HELD(phm)); panic("page_lookup_create: hashin failed %p %p %llx %p", (void *)newpp, (void *)vp, off, (void *)phm); /*NOTREACHED*/ } ASSERT(MUTEX_HELD(phm)); mutex_exit(phm); phm = NULL; page_set_props(newpp, P_REF); page_io_lock(newpp); pp = newpp; se = SE_EXCL; } else { VM_STAT_ADD(page_lookup_cnt[19]); mutex_exit(phm); } ASSERT(pp ? PAGE_LOCKED_SE(pp, se) : 1); ASSERT(pp ? ((PP_ISFREE(pp) == 0) && (PP_ISAGED(pp) == 0)) : 1); return (pp); } /* * Search the hash list for the page representing the * specified [vp, offset] and return it locked. Skip * free pages and pages that cannot be locked as requested. * Used while attempting to kluster pages. */ page_t * page_lookup_nowait(vnode_t *vp, u_offset_t off, se_t se) { page_t *pp; kmutex_t *phm; ulong_t index; uint_t locked; ASSERT(MUTEX_NOT_HELD(page_vnode_mutex(vp))); VM_STAT_ADD(page_lookup_nowait_cnt[0]); index = PAGE_HASH_FUNC(vp, off); pp = page_hash_search(index, vp, off); locked = 0; if (pp == NULL) { top: VM_STAT_ADD(page_lookup_nowait_cnt[1]); locked = 1; phm = PAGE_HASH_MUTEX(index); mutex_enter(phm); pp = page_hash_search(index, vp, off); } if (pp == NULL || PP_ISFREE(pp)) { VM_STAT_ADD(page_lookup_nowait_cnt[2]); pp = NULL; } else { if (!page_trylock(pp, se)) { VM_STAT_ADD(page_lookup_nowait_cnt[3]); pp = NULL; } else { VM_STAT_ADD(page_lookup_nowait_cnt[4]); /* * See the comment in page_lookup() */ if (((volatile struct vnode *)(pp->p_vnode) != vp) || ((u_offset_t)(pp->p_offset) != off)) { VM_STAT_ADD(page_lookup_nowait_cnt[5]); if (locked) { panic("page_lookup_nowait %p", (void *)pp); /*NOTREACHED*/ } page_unlock(pp); goto top; } if (PP_ISFREE(pp)) { VM_STAT_ADD(page_lookup_nowait_cnt[6]); page_unlock(pp); pp = NULL; } } } if (locked) { VM_STAT_ADD(page_lookup_nowait_cnt[7]); mutex_exit(phm); } ASSERT(pp ? PAGE_LOCKED_SE(pp, se) : 1); return (pp); } /* * Search the hash list for a page with the specified [vp, off] * that is known to exist and is already locked. This routine * is typically used by segment SOFTUNLOCK routines. */ page_t * page_find(vnode_t *vp, u_offset_t off) { page_t *pp; kmutex_t *phm; ulong_t index; ASSERT(MUTEX_NOT_HELD(page_vnode_mutex(vp))); VM_STAT_ADD(page_find_cnt); index = PAGE_HASH_FUNC(vp, off); phm = PAGE_HASH_MUTEX(index); mutex_enter(phm); pp = page_hash_search(index, vp, off); mutex_exit(phm); ASSERT(pp == NULL || PAGE_LOCKED(pp) || panicstr); return (pp); } /* * Determine whether a page with the specified [vp, off] * currently exists in the system. Obviously this should * only be considered as a hint since nothing prevents the * page from disappearing or appearing immediately after * the return from this routine. Subsequently, we don't * even bother to lock the list. */ page_t * page_exists(vnode_t *vp, u_offset_t off) { ulong_t index; ASSERT(MUTEX_NOT_HELD(page_vnode_mutex(vp))); VM_STAT_ADD(page_exists_cnt); index = PAGE_HASH_FUNC(vp, off); return (page_hash_search(index, vp, off)); } /* * Determine if physically contiguous pages exist for [vp, off] - [vp, off + * page_size(szc)) range. if they exist and ppa is not NULL fill ppa array * with these pages locked SHARED. If necessary reclaim pages from * freelist. Return 1 if contiguous pages exist and 0 otherwise. * * If we fail to lock pages still return 1 if pages exist and contiguous. * But in this case return value is just a hint. ppa array won't be filled. * Caller should initialize ppa[0] as NULL to distinguish return value. * * Returns 0 if pages don't exist or not physically contiguous. * * This routine doesn't work for anonymous(swapfs) pages. */ int page_exists_physcontig(vnode_t *vp, u_offset_t off, uint_t szc, page_t *ppa[]) { pgcnt_t pages; pfn_t pfn; page_t *rootpp; pgcnt_t i; pgcnt_t j; u_offset_t save_off = off; ulong_t index; kmutex_t *phm; page_t *pp; uint_t pszc; int loopcnt = 0; ASSERT(szc != 0); ASSERT(vp != NULL); ASSERT(!IS_SWAPFSVP(vp)); ASSERT(!VN_ISKAS(vp)); again: if (++loopcnt > 3) { VM_STAT_ADD(page_exphcontg[0]); return (0); } index = PAGE_HASH_FUNC(vp, off); phm = PAGE_HASH_MUTEX(index); mutex_enter(phm); pp = page_hash_search(index, vp, off); mutex_exit(phm); VM_STAT_ADD(page_exphcontg[1]); if (pp == NULL) { VM_STAT_ADD(page_exphcontg[2]); return (0); } pages = page_get_pagecnt(szc); rootpp = pp; pfn = rootpp->p_pagenum; if ((pszc = pp->p_szc) >= szc && ppa != NULL) { VM_STAT_ADD(page_exphcontg[3]); if (!page_trylock(pp, SE_SHARED)) { VM_STAT_ADD(page_exphcontg[4]); return (1); } /* * Also check whether p_pagenum was modified by DR. */ if (pp->p_szc != pszc || pp->p_vnode != vp || pp->p_offset != off || pp->p_pagenum != pfn) { VM_STAT_ADD(page_exphcontg[5]); page_unlock(pp); off = save_off; goto again; } /* * szc was non zero and vnode and offset matched after we * locked the page it means it can't become free on us. */ ASSERT(!PP_ISFREE(pp)); if (!IS_P2ALIGNED(pfn, pages)) { page_unlock(pp); return (0); } ppa[0] = pp; pp++; off += PAGESIZE; pfn++; for (i = 1; i < pages; i++, pp++, off += PAGESIZE, pfn++) { if (!page_trylock(pp, SE_SHARED)) { VM_STAT_ADD(page_exphcontg[6]); pp--; while (i-- > 0) { page_unlock(pp); pp--; } ppa[0] = NULL; return (1); } if (pp->p_szc != pszc) { VM_STAT_ADD(page_exphcontg[7]); page_unlock(pp); pp--; while (i-- > 0) { page_unlock(pp); pp--; } ppa[0] = NULL; off = save_off; goto again; } /* * szc the same as for previous already locked pages * with right identity. Since this page had correct * szc after we locked it can't get freed or destroyed * and therefore must have the expected identity. */ ASSERT(!PP_ISFREE(pp)); if (pp->p_vnode != vp || pp->p_offset != off) { panic("page_exists_physcontig: " "large page identity doesn't match"); } ppa[i] = pp; ASSERT(pp->p_pagenum == pfn); } VM_STAT_ADD(page_exphcontg[8]); ppa[pages] = NULL; return (1); } else if (pszc >= szc) { VM_STAT_ADD(page_exphcontg[9]); if (!IS_P2ALIGNED(pfn, pages)) { return (0); } return (1); } if (!IS_P2ALIGNED(pfn, pages)) { VM_STAT_ADD(page_exphcontg[10]); return (0); } if (page_numtomemseg_nolock(pfn) != page_numtomemseg_nolock(pfn + pages - 1)) { VM_STAT_ADD(page_exphcontg[11]); return (0); } /* * We loop up 4 times across pages to promote page size. * We're extra cautious to promote page size atomically with respect * to everybody else. But we can probably optimize into 1 loop if * this becomes an issue. */ for (i = 0; i < pages; i++, pp++, off += PAGESIZE, pfn++) { if (!page_trylock(pp, SE_EXCL)) { VM_STAT_ADD(page_exphcontg[12]); break; } /* * Check whether p_pagenum was modified by DR. */ if (pp->p_pagenum != pfn) { page_unlock(pp); break; } if (pp->p_vnode != vp || pp->p_offset != off) { VM_STAT_ADD(page_exphcontg[13]); page_unlock(pp); break; } if (pp->p_szc >= szc) { ASSERT(i == 0); page_unlock(pp); off = save_off; goto again; } } if (i != pages) { VM_STAT_ADD(page_exphcontg[14]); --pp; while (i-- > 0) { page_unlock(pp); --pp; } return (0); } pp = rootpp; for (i = 0; i < pages; i++, pp++) { if (PP_ISFREE(pp)) { VM_STAT_ADD(page_exphcontg[15]); ASSERT(!PP_ISAGED(pp)); ASSERT(pp->p_szc == 0); if (!page_reclaim(pp, NULL)) { break; } } else { ASSERT(pp->p_szc < szc); VM_STAT_ADD(page_exphcontg[16]); (void) hat_pageunload(pp, HAT_FORCE_PGUNLOAD); } } if (i < pages) { VM_STAT_ADD(page_exphcontg[17]); /* * page_reclaim failed because we were out of memory. * drop the rest of the locks and return because this page * must be already reallocated anyway. */ pp = rootpp; for (j = 0; j < pages; j++, pp++) { if (j != i) { page_unlock(pp); } } return (0); } off = save_off; pp = rootpp; for (i = 0; i < pages; i++, pp++, off += PAGESIZE) { ASSERT(PAGE_EXCL(pp)); ASSERT(!PP_ISFREE(pp)); ASSERT(!hat_page_is_mapped(pp)); ASSERT(pp->p_vnode == vp); ASSERT(pp->p_offset == off); pp->p_szc = szc; } pp = rootpp; for (i = 0; i < pages; i++, pp++) { if (ppa == NULL) { page_unlock(pp); } else { ppa[i] = pp; page_downgrade(ppa[i]); } } if (ppa != NULL) { ppa[pages] = NULL; } VM_STAT_ADD(page_exphcontg[18]); ASSERT(vp->v_pages != NULL); return (1); } /* * Determine whether a page with the specified [vp, off] * currently exists in the system and if so return its * size code. Obviously this should only be considered as * a hint since nothing prevents the page from disappearing * or appearing immediately after the return from this routine. */ int page_exists_forreal(vnode_t *vp, u_offset_t off, uint_t *szc) { page_t *pp; kmutex_t *phm; ulong_t index; int rc = 0; ASSERT(MUTEX_NOT_HELD(page_vnode_mutex(vp))); ASSERT(szc != NULL); VM_STAT_ADD(page_exists_forreal_cnt); index = PAGE_HASH_FUNC(vp, off); phm = PAGE_HASH_MUTEX(index); mutex_enter(phm); pp = page_hash_search(index, vp, off); if (pp != NULL) { *szc = pp->p_szc; rc = 1; } mutex_exit(phm); return (rc); } /* wakeup threads waiting for pages in page_create_get_something() */ void wakeup_pcgs(void) { if (!CV_HAS_WAITERS(&pcgs_cv)) return; cv_broadcast(&pcgs_cv); } /* * 'freemem' is used all over the kernel as an indication of how many * pages are free (either on the cache list or on the free page list) * in the system. In very few places is a really accurate 'freemem' * needed. To avoid contention of the lock protecting a the * single freemem, it was spread out into NCPU buckets. Set_freemem * sets freemem to the total of all NCPU buckets. It is called from * clock() on each TICK. */ void set_freemem(void) { struct pcf *p; ulong_t t; uint_t i; t = 0; p = pcf; for (i = 0; i < pcf_fanout; i++) { t += p->pcf_count; p++; } freemem = t; /* * Don't worry about grabbing mutex. It's not that * critical if we miss a tick or two. This is * where we wakeup possible delayers in * page_create_get_something(). */ wakeup_pcgs(); } ulong_t get_freemem() { struct pcf *p; ulong_t t; uint_t i; t = 0; p = pcf; for (i = 0; i < pcf_fanout; i++) { t += p->pcf_count; p++; } /* * We just calculated it, might as well set it. */ freemem = t; return (t); } /* * Acquire all of the page cache & free (pcf) locks. */ void pcf_acquire_all() { struct pcf *p; uint_t i; p = pcf; for (i = 0; i < pcf_fanout; i++) { mutex_enter(&p->pcf_lock); p++; } } /* * Release all the pcf_locks. */ void pcf_release_all() { struct pcf *p; uint_t i; p = pcf; for (i = 0; i < pcf_fanout; i++) { mutex_exit(&p->pcf_lock); p++; } } /* * Inform the VM system that we need some pages freed up. * Calls must be symmetric, e.g.: * * page_needfree(100); * wait a bit; * page_needfree(-100); */ void page_needfree(spgcnt_t npages) { mutex_enter(&new_freemem_lock); needfree += npages; mutex_exit(&new_freemem_lock); } /* * Throttle for page_create(): try to prevent freemem from dropping * below throttlefree. We can't provide a 100% guarantee because * KM_NOSLEEP allocations, page_reclaim(), and various other things * nibble away at the freelist. However, we can block all PG_WAIT * allocations until memory becomes available. The motivation is * that several things can fall apart when there's no free memory: * * (1) If pageout() needs memory to push a page, the system deadlocks. * * (2) By (broken) specification, timeout(9F) can neither fail nor * block, so it has no choice but to panic the system if it * cannot allocate a callout structure. * * (3) Like timeout(), ddi_set_callback() cannot fail and cannot block; * it panics if it cannot allocate a callback structure. * * (4) Untold numbers of third-party drivers have not yet been hardened * against KM_NOSLEEP and/or allocb() failures; they simply assume * success and panic the system with a data fault on failure. * (The long-term solution to this particular problem is to ship * hostile fault-injecting DEBUG kernels with the DDK.) * * It is theoretically impossible to guarantee success of non-blocking * allocations, but in practice, this throttle is very hard to break. */ static int page_create_throttle(pgcnt_t npages, int flags) { ulong_t fm; uint_t i; pgcnt_t tf; /* effective value of throttlefree */ atomic_inc_64(&n_throttle); /* * Normal priority allocations. */ if ((flags & (PG_WAIT | PG_NORMALPRI)) == PG_NORMALPRI) { ASSERT(!(flags & (PG_PANIC | PG_PUSHPAGE))); return (freemem >= npages + throttlefree); } /* * Never deny pages when: * - it's a thread that cannot block [NOMEMWAIT()] * - the allocation cannot block and must not fail * - the allocation cannot block and is pageout dispensated */ if (NOMEMWAIT() || ((flags & (PG_WAIT | PG_PANIC)) == PG_PANIC) || ((flags & (PG_WAIT | PG_PUSHPAGE)) == PG_PUSHPAGE)) return (1); /* * If the allocation can't block, we look favorably upon it * unless we're below pageout_reserve. In that case we fail * the allocation because we want to make sure there are a few * pages available for pageout. */ if ((flags & PG_WAIT) == 0) return (freemem >= npages + pageout_reserve); /* Calculate the effective throttlefree value */ tf = throttlefree - ((flags & PG_PUSHPAGE) ? pageout_reserve : 0); WAKE_PAGEOUT_SCANNER(page__create__throttle); for (;;) { fm = 0; pcf_acquire_all(); mutex_enter(&new_freemem_lock); for (i = 0; i < pcf_fanout; i++) { fm += pcf[i].pcf_count; pcf[i].pcf_wait++; mutex_exit(&pcf[i].pcf_lock); } freemem = fm; if (freemem >= npages + tf) { mutex_exit(&new_freemem_lock); break; } needfree += npages; freemem_wait++; cv_wait(&freemem_cv, &new_freemem_lock); freemem_wait--; needfree -= npages; mutex_exit(&new_freemem_lock); } return (1); } /* * page_create_wait() is called to either coalesce pages from the * different pcf buckets or to wait because there simply are not * enough pages to satisfy the caller's request. * * Sadly, this is called from platform/vm/vm_machdep.c */ int page_create_wait(pgcnt_t npages, uint_t flags) { pgcnt_t total; uint_t i; struct pcf *p; /* * Wait until there are enough free pages to satisfy our * entire request. * We set needfree += npages before prodding pageout, to make sure * it does real work when npages > lotsfree > freemem. */ VM_STAT_ADD(page_create_not_enough); ASSERT(!kcage_on ? !(flags & PG_NORELOC) : 1); checkagain: if ((flags & PG_NORELOC) && kcage_freemem < kcage_throttlefree + npages) (void) kcage_create_throttle(npages, flags); if (freemem < npages + throttlefree) if (!page_create_throttle(npages, flags)) return (0); if (pcf_decrement_bucket(npages) || pcf_decrement_multiple(&total, npages, 0)) return (1); /* * All of the pcf locks are held, there are not enough pages * to satisfy the request (npages < total). * Be sure to acquire the new_freemem_lock before dropping * the pcf locks. This prevents dropping wakeups in page_free(). * The order is always pcf_lock then new_freemem_lock. * * Since we hold all the pcf locks, it is a good time to set freemem. * * If the caller does not want to wait, return now. * Else turn the pageout daemon loose to find something * and wait till it does. * */ freemem = total; if ((flags & PG_WAIT) == 0) { pcf_release_all(); TRACE_2(TR_FAC_VM, TR_PAGE_CREATE_NOMEM, "page_create_nomem:npages %ld freemem %ld", npages, freemem); return (0); } ASSERT(proc_pageout != NULL); WAKE_PAGEOUT_SCANNER(page__create__wait); TRACE_2(TR_FAC_VM, TR_PAGE_CREATE_SLEEP_START, "page_create_sleep_start: freemem %ld needfree %ld", freemem, needfree); /* * We are going to wait. * We currently hold all of the pcf_locks, * get the new_freemem_lock (it protects freemem_wait), * before dropping the pcf_locks. */ mutex_enter(&new_freemem_lock); p = pcf; for (i = 0; i < pcf_fanout; i++) { p->pcf_wait++; mutex_exit(&p->pcf_lock); p++; } needfree += npages; freemem_wait++; cv_wait(&freemem_cv, &new_freemem_lock); freemem_wait--; needfree -= npages; mutex_exit(&new_freemem_lock); TRACE_2(TR_FAC_VM, TR_PAGE_CREATE_SLEEP_END, "page_create_sleep_end: freemem %ld needfree %ld", freemem, needfree); VM_STAT_ADD(page_create_not_enough_again); goto checkagain; } /* * A routine to do the opposite of page_create_wait(). */ void page_create_putback(spgcnt_t npages) { struct pcf *p; pgcnt_t lump; uint_t *which; /* * When a contiguous lump is broken up, we have to * deal with lots of pages (min 64) so lets spread * the wealth around. */ lump = roundup(npages, pcf_fanout) / pcf_fanout; freemem += npages; for (p = pcf; (npages > 0) && (p < &pcf[pcf_fanout]); p++) { which = &p->pcf_count; mutex_enter(&p->pcf_lock); if (p->pcf_block) { which = &p->pcf_reserve; } if (lump < npages) { *which += (uint_t)lump; npages -= lump; } else { *which += (uint_t)npages; npages = 0; } if (p->pcf_wait) { mutex_enter(&new_freemem_lock); /* * Check to see if some other thread * is actually waiting. Another bucket * may have woken it up by now. If there * are no waiters, then set our pcf_wait * count to zero to avoid coming in here * next time. */ if (freemem_wait) { if (npages > 1) { cv_broadcast(&freemem_cv); } else { cv_signal(&freemem_cv); } p->pcf_wait--; } else { p->pcf_wait = 0; } mutex_exit(&new_freemem_lock); } mutex_exit(&p->pcf_lock); } ASSERT(npages == 0); } /* * A helper routine for page_create_get_something. * The indenting got to deep down there. * Unblock the pcf counters. Any pages freed after * pcf_block got set are moved to pcf_count and * wakeups (cv_broadcast() or cv_signal()) are done as needed. */ static void pcgs_unblock(void) { int i; struct pcf *p; /* Update freemem while we're here. */ freemem = 0; p = pcf; for (i = 0; i < pcf_fanout; i++) { mutex_enter(&p->pcf_lock); ASSERT(p->pcf_count == 0); p->pcf_count = p->pcf_reserve; p->pcf_block = 0; freemem += p->pcf_count; if (p->pcf_wait) { mutex_enter(&new_freemem_lock); if (freemem_wait) { if (p->pcf_reserve > 1) { cv_broadcast(&freemem_cv); p->pcf_wait = 0; } else { cv_signal(&freemem_cv); p->pcf_wait--; } } else { p->pcf_wait = 0; } mutex_exit(&new_freemem_lock); } p->pcf_reserve = 0; mutex_exit(&p->pcf_lock); p++; } } /* * Called from page_create_va() when both the cache and free lists * have been checked once. * * Either returns a page or panics since the accounting was done * way before we got here. * * We don't come here often, so leave the accounting on permanently. */ #define MAX_PCGS 100 #ifdef DEBUG #define PCGS_TRIES 100 #else /* DEBUG */ #define PCGS_TRIES 10 #endif /* DEBUG */ #ifdef VM_STATS uint_t pcgs_counts[PCGS_TRIES]; uint_t pcgs_too_many; uint_t pcgs_entered; uint_t pcgs_entered_noreloc; uint_t pcgs_locked; uint_t pcgs_cagelocked; #endif /* VM_STATS */ static page_t * page_create_get_something(vnode_t *vp, u_offset_t off, struct seg *seg, caddr_t vaddr, uint_t flags) { uint_t count; page_t *pp; uint_t locked, i; struct pcf *p; lgrp_t *lgrp; int cagelocked = 0; VM_STAT_ADD(pcgs_entered); /* * Tap any reserve freelists: if we fail now, we'll die * since the page(s) we're looking for have already been * accounted for. */ flags |= PG_PANIC; if ((flags & PG_NORELOC) != 0) { VM_STAT_ADD(pcgs_entered_noreloc); /* * Requests for free pages from critical threads * such as pageout still won't throttle here, but * we must try again, to give the cageout thread * another chance to catch up. Since we already * accounted for the pages, we had better get them * this time. * * N.B. All non-critical threads acquire the pcgs_cagelock * to serialize access to the freelists. This implements a * turnstile-type synchornization to avoid starvation of * critical requests for PG_NORELOC memory by non-critical * threads: all non-critical threads must acquire a 'ticket' * before passing through, which entails making sure * kcage_freemem won't fall below minfree prior to grabbing * pages from the freelists. */ if (kcage_create_throttle(1, flags) == KCT_NONCRIT) { mutex_enter(&pcgs_cagelock); cagelocked = 1; VM_STAT_ADD(pcgs_cagelocked); } } /* * Time to get serious. * We failed to get a `correctly colored' page from both the * free and cache lists. * We escalate in stage. * * First try both lists without worring about color. * * Then, grab all page accounting locks (ie. pcf[]) and * steal any pages that they have and set the pcf_block flag to * stop deletions from the lists. This will help because * a page can get added to the free list while we are looking * at the cache list, then another page could be added to the cache * list allowing the page on the free list to be removed as we * move from looking at the cache list to the free list. This * could happen over and over. We would never find the page * we have accounted for. * * Noreloc pages are a subset of the global (relocatable) page pool. * They are not tracked separately in the pcf bins, so it is * impossible to know when doing pcf accounting if the available * page(s) are noreloc pages or not. When looking for a noreloc page * it is quite easy to end up here even if the global (relocatable) * page pool has plenty of free pages but the noreloc pool is empty. * * When the noreloc pool is empty (or low), additional noreloc pages * are created by converting pages from the global page pool. This * process will stall during pcf accounting if the pcf bins are * already locked. Such is the case when a noreloc allocation is * looping here in page_create_get_something waiting for more noreloc * pages to appear. * * Short of adding a new field to the pcf bins to accurately track * the number of free noreloc pages, we instead do not grab the * pcgs_lock, do not set the pcf blocks and do not timeout when * allocating a noreloc page. This allows noreloc allocations to * loop without blocking global page pool allocations. * * NOTE: the behaviour of page_create_get_something has not changed * for the case of global page pool allocations. */ flags &= ~PG_MATCH_COLOR; locked = 0; #if defined(__x86) flags = page_create_update_flags_x86(flags); #endif lgrp = lgrp_mem_choose(seg, vaddr, PAGESIZE); for (count = 0; kcage_on || count < MAX_PCGS; count++) { pp = page_get_freelist(vp, off, seg, vaddr, PAGESIZE, flags, lgrp); if (pp == NULL) { pp = page_get_cachelist(vp, off, seg, vaddr, flags, lgrp); } if (pp == NULL) { /* * Serialize. Don't fight with other pcgs(). */ if (!locked && (!kcage_on || !(flags & PG_NORELOC))) { mutex_enter(&pcgs_lock); VM_STAT_ADD(pcgs_locked); locked = 1; p = pcf; for (i = 0; i < pcf_fanout; i++) { mutex_enter(&p->pcf_lock); ASSERT(p->pcf_block == 0); p->pcf_block = 1; p->pcf_reserve = p->pcf_count; p->pcf_count = 0; mutex_exit(&p->pcf_lock); p++; } freemem = 0; } if (count) { /* * Since page_free() puts pages on * a list then accounts for it, we * just have to wait for page_free() * to unlock any page it was working * with. The page_lock()-page_reclaim() * path falls in the same boat. * * We don't need to check on the * PG_WAIT flag, we have already * accounted for the page we are * looking for in page_create_va(). * * We just wait a moment to let any * locked pages on the lists free up, * then continue around and try again. * * Will be awakened by set_freemem(). */ mutex_enter(&pcgs_wait_lock); cv_wait(&pcgs_cv, &pcgs_wait_lock); mutex_exit(&pcgs_wait_lock); } } else { #ifdef VM_STATS if (count >= PCGS_TRIES) { VM_STAT_ADD(pcgs_too_many); } else { VM_STAT_ADD(pcgs_counts[count]); } #endif if (locked) { pcgs_unblock(); mutex_exit(&pcgs_lock); } if (cagelocked) mutex_exit(&pcgs_cagelock); return (pp); } } /* * we go down holding the pcf locks. */ panic("no %spage found %d", ((flags & PG_NORELOC) ? "non-reloc " : ""), count); /*NOTREACHED*/ } /* * Create enough pages for "bytes" worth of data starting at * "off" in "vp". * * Where flag must be one of: * * PG_EXCL: Exclusive create (fail if any page already * exists in the page cache) which does not * wait for memory to become available. * * PG_WAIT: Non-exclusive create which can wait for * memory to become available. * * PG_PHYSCONTIG: Allocate physically contiguous pages. * (Not Supported) * * A doubly linked list of pages is returned to the caller. Each page * on the list has the "exclusive" (p_selock) lock and "iolock" (p_iolock) * lock. * * Unable to change the parameters to page_create() in a minor release, * we renamed page_create() to page_create_va(), changed all known calls * from page_create() to page_create_va(), and created this wrapper. * * Upon a major release, we should break compatibility by deleting this * wrapper, and replacing all the strings "page_create_va", with "page_create". * * NOTE: There is a copy of this interface as page_create_io() in * i86/vm/vm_machdep.c. Any bugs fixed here should be applied * there. */ page_t * page_create(vnode_t *vp, u_offset_t off, size_t bytes, uint_t flags) { caddr_t random_vaddr; struct seg kseg; #ifdef DEBUG cmn_err(CE_WARN, "Using deprecated interface page_create: caller %p", (void *)caller()); #endif random_vaddr = (caddr_t)(((uintptr_t)vp >> 7) ^ (uintptr_t)(off >> PAGESHIFT)); kseg.s_as = &kas; return (page_create_va(vp, off, bytes, flags, &kseg, random_vaddr)); } #ifdef DEBUG uint32_t pg_alloc_pgs_mtbf = 0; #endif /* * Used for large page support. It will attempt to allocate * a large page(s) off the freelist. * * Returns non zero on failure. */ int page_alloc_pages(struct vnode *vp, struct seg *seg, caddr_t addr, page_t **basepp, page_t *ppa[], uint_t szc, int anypgsz, int pgflags) { pgcnt_t npgs, curnpgs, totpgs; size_t pgsz; page_t *pplist = NULL, *pp; int err = 0; lgrp_t *lgrp; ASSERT(szc != 0 && szc <= (page_num_pagesizes() - 1)); ASSERT(pgflags == 0 || pgflags == PG_LOCAL); /* * Check if system heavily prefers local large pages over remote * on systems with multiple lgroups. */ if (lpg_alloc_prefer == LPAP_LOCAL && nlgrps > 1) { pgflags = PG_LOCAL; } VM_STAT_ADD(alloc_pages[0]); #ifdef DEBUG if (pg_alloc_pgs_mtbf && !(gethrtime() % pg_alloc_pgs_mtbf)) { return (ENOMEM); } #endif /* * One must be NULL but not both. * And one must be non NULL but not both. */ ASSERT(basepp != NULL || ppa != NULL); ASSERT(basepp == NULL || ppa == NULL); #if defined(__x86) while (page_chk_freelist(szc) == 0) { VM_STAT_ADD(alloc_pages[8]); if (anypgsz == 0 || --szc == 0) return (ENOMEM); } #endif pgsz = page_get_pagesize(szc); totpgs = curnpgs = npgs = pgsz >> PAGESHIFT; ASSERT(((uintptr_t)addr & (pgsz - 1)) == 0); (void) page_create_wait(npgs, PG_WAIT); while (npgs && szc) { lgrp = lgrp_mem_choose(seg, addr, pgsz); if (pgflags == PG_LOCAL) { pp = page_get_freelist(vp, 0, seg, addr, pgsz, pgflags, lgrp); if (pp == NULL) { pp = page_get_freelist(vp, 0, seg, addr, pgsz, 0, lgrp); } } else { pp = page_get_freelist(vp, 0, seg, addr, pgsz, 0, lgrp); } if (pp != NULL) { VM_STAT_ADD(alloc_pages[1]); page_list_concat(&pplist, &pp); ASSERT(npgs >= curnpgs); npgs -= curnpgs; } else if (anypgsz) { VM_STAT_ADD(alloc_pages[2]); szc--; pgsz = page_get_pagesize(szc); curnpgs = pgsz >> PAGESHIFT; } else { VM_STAT_ADD(alloc_pages[3]); ASSERT(npgs == totpgs); page_create_putback(npgs); return (ENOMEM); } } if (szc == 0) { VM_STAT_ADD(alloc_pages[4]); ASSERT(npgs != 0); page_create_putback(npgs); err = ENOMEM; } else if (basepp != NULL) { ASSERT(npgs == 0); ASSERT(ppa == NULL); *basepp = pplist; } npgs = totpgs - npgs; pp = pplist; /* * Clear the free and age bits. Also if we were passed in a ppa then * fill it in with all the constituent pages from the large page. But * if we failed to allocate all the pages just free what we got. */ while (npgs != 0) { ASSERT(PP_ISFREE(pp)); ASSERT(PP_ISAGED(pp)); if (ppa != NULL || err != 0) { if (err == 0) { VM_STAT_ADD(alloc_pages[5]); PP_CLRFREE(pp); PP_CLRAGED(pp); page_sub(&pplist, pp); *ppa++ = pp; npgs--; } else { VM_STAT_ADD(alloc_pages[6]); ASSERT(pp->p_szc != 0); curnpgs = page_get_pagecnt(pp->p_szc); page_list_break(&pp, &pplist, curnpgs); page_list_add_pages(pp, 0); page_create_putback(curnpgs); ASSERT(npgs >= curnpgs); npgs -= curnpgs; } pp = pplist; } else { VM_STAT_ADD(alloc_pages[7]); PP_CLRFREE(pp); PP_CLRAGED(pp); pp = pp->p_next; npgs--; } } return (err); } /* * Get a single large page off of the freelists, and set it up for use. * Number of bytes requested must be a supported page size. * * Note that this call may fail even if there is sufficient * memory available or PG_WAIT is set, so the caller must * be willing to fallback on page_create_va(), block and retry, * or fail the requester. */ page_t * page_create_va_large(vnode_t *vp, u_offset_t off, size_t bytes, uint_t flags, struct seg *seg, caddr_t vaddr, void *arg) { pgcnt_t npages; page_t *pp; page_t *rootpp; lgrp_t *lgrp; lgrp_id_t *lgrpid = (lgrp_id_t *)arg; ASSERT(vp != NULL); ASSERT((flags & ~(PG_EXCL | PG_WAIT | PG_NORELOC | PG_PANIC | PG_PUSHPAGE | PG_NORMALPRI)) == 0); /* but no others */ ASSERT((flags & PG_EXCL) == PG_EXCL); npages = btop(bytes); if (!kcage_on || panicstr) { /* * Cage is OFF, or we are single threaded in * panic, so make everything a RELOC request. */ flags &= ~PG_NORELOC; } /* * Make sure there's adequate physical memory available. * Note: PG_WAIT is ignored here. */ if (freemem <= throttlefree + npages) { VM_STAT_ADD(page_create_large_cnt[1]); return (NULL); } /* * If cage is on, dampen draw from cage when available * cage space is low. */ if ((flags & (PG_NORELOC | PG_WAIT)) == (PG_NORELOC | PG_WAIT) && kcage_freemem < kcage_throttlefree + npages) { /* * The cage is on, the caller wants PG_NORELOC * pages and available cage memory is very low. * Call kcage_create_throttle() to attempt to * control demand on the cage. */ if (kcage_create_throttle(npages, flags) == KCT_FAILURE) { VM_STAT_ADD(page_create_large_cnt[2]); return (NULL); } } if (!pcf_decrement_bucket(npages) && !pcf_decrement_multiple(NULL, npages, 1)) { VM_STAT_ADD(page_create_large_cnt[4]); return (NULL); } /* * This is where this function behaves fundamentally differently * than page_create_va(); since we're intending to map the page * with a single TTE, we have to get it as a physically contiguous * hardware pagesize chunk. If we can't, we fail. */ if (lgrpid != NULL && *lgrpid >= 0 && *lgrpid <= lgrp_alloc_max && LGRP_EXISTS(lgrp_table[*lgrpid])) lgrp = lgrp_table[*lgrpid]; else lgrp = lgrp_mem_choose(seg, vaddr, bytes); if ((rootpp = page_get_freelist(&kvp, off, seg, vaddr, bytes, flags & ~PG_MATCH_COLOR, lgrp)) == NULL) { page_create_putback(npages); VM_STAT_ADD(page_create_large_cnt[5]); return (NULL); } /* * if we got the page with the wrong mtype give it back this is a * workaround for CR 6249718. When CR 6249718 is fixed we never get * inside "if" and the workaround becomes just a nop */ if (kcage_on && (flags & PG_NORELOC) && !PP_ISNORELOC(rootpp)) { page_list_add_pages(rootpp, 0); page_create_putback(npages); VM_STAT_ADD(page_create_large_cnt[6]); return (NULL); } /* * If satisfying this request has left us with too little * memory, start the wheels turning to get some back. The * first clause of the test prevents waking up the pageout * daemon in situations where it would decide that there's * nothing to do. */ if (nscan < desscan && freemem < minfree) { TRACE_1(TR_FAC_VM, TR_PAGEOUT_CV_SIGNAL, "pageout_cv_signal:freemem %ld", freemem); WAKE_PAGEOUT_SCANNER(va__large); } pp = rootpp; while (npages--) { ASSERT(PAGE_EXCL(pp)); ASSERT(pp->p_vnode == NULL); ASSERT(!hat_page_is_mapped(pp)); PP_CLRFREE(pp); PP_CLRAGED(pp); if (!page_hashin(pp, vp, off, NULL)) panic("page_create_large: hashin failed: page %p", (void *)pp); page_io_lock(pp); off += PAGESIZE; pp = pp->p_next; } VM_STAT_ADD(page_create_large_cnt[0]); return (rootpp); } page_t * page_create_va(vnode_t *vp, u_offset_t off, size_t bytes, uint_t flags, struct seg *seg, caddr_t vaddr) { page_t *plist = NULL; pgcnt_t npages; pgcnt_t found_on_free = 0; pgcnt_t pages_req; page_t *npp = NULL; struct pcf *p; lgrp_t *lgrp; TRACE_4(TR_FAC_VM, TR_PAGE_CREATE_START, "page_create_start:vp %p off %llx bytes %lu flags %x", vp, off, bytes, flags); ASSERT(bytes != 0 && vp != NULL); if ((flags & PG_EXCL) == 0 && (flags & PG_WAIT) == 0) { panic("page_create: invalid flags"); /*NOTREACHED*/ } ASSERT((flags & ~(PG_EXCL | PG_WAIT | PG_NORELOC | PG_PANIC | PG_PUSHPAGE | PG_NORMALPRI)) == 0); /* but no others */ pages_req = npages = btopr(bytes); /* * Try to see whether request is too large to *ever* be * satisfied, in order to prevent deadlock. We arbitrarily * decide to limit maximum size requests to max_page_get. */ if (npages >= max_page_get) { if ((flags & PG_WAIT) == 0) { TRACE_4(TR_FAC_VM, TR_PAGE_CREATE_TOOBIG, "page_create_toobig:vp %p off %llx npages " "%lu max_page_get %lu", vp, off, npages, max_page_get); return (NULL); } else { cmn_err(CE_WARN, "Request for too much kernel memory " "(%lu bytes), will hang forever", bytes); for (;;) delay(1000000000); } } if (!kcage_on || panicstr) { /* * Cage is OFF, or we are single threaded in * panic, so make everything a RELOC request. */ flags &= ~PG_NORELOC; } if (freemem <= throttlefree + npages) if (!page_create_throttle(npages, flags)) return (NULL); /* * If cage is on, dampen draw from cage when available * cage space is low. */ if ((flags & PG_NORELOC) && kcage_freemem < kcage_throttlefree + npages) { /* * The cage is on, the caller wants PG_NORELOC * pages and available cage memory is very low. * Call kcage_create_throttle() to attempt to * control demand on the cage. */ if (kcage_create_throttle(npages, flags) == KCT_FAILURE) return (NULL); } VM_STAT_ADD(page_create_cnt[0]); if (!pcf_decrement_bucket(npages)) { /* * Have to look harder. If npages is greater than * one, then we might have to coalesce the counters. * * Go wait. We come back having accounted * for the memory. */ VM_STAT_ADD(page_create_cnt[1]); if (!page_create_wait(npages, flags)) { VM_STAT_ADD(page_create_cnt[2]); return (NULL); } } TRACE_2(TR_FAC_VM, TR_PAGE_CREATE_SUCCESS, "page_create_success:vp %p off %llx", vp, off); /* * If satisfying this request has left us with too little * memory, start the wheels turning to get some back. The * first clause of the test prevents waking up the pageout * daemon in situations where it would decide that there's * nothing to do. */ if (nscan < desscan && freemem < minfree) { TRACE_1(TR_FAC_VM, TR_PAGEOUT_CV_SIGNAL, "pageout_cv_signal:freemem %ld", freemem); WAKE_PAGEOUT_SCANNER(va); } /* * Loop around collecting the requested number of pages. * Most of the time, we have to `create' a new page. With * this in mind, pull the page off the free list before * getting the hash lock. This will minimize the hash * lock hold time, nesting, and the like. If it turns * out we don't need the page, we put it back at the end. */ while (npages--) { page_t *pp; kmutex_t *phm = NULL; ulong_t index; index = PAGE_HASH_FUNC(vp, off); top: ASSERT(phm == NULL); ASSERT(index == PAGE_HASH_FUNC(vp, off)); ASSERT(MUTEX_NOT_HELD(page_vnode_mutex(vp))); if (npp == NULL) { /* * Try to get a page from the freelist (ie, * a page with no [vp, off] tag). If that * fails, use the cachelist. * * During the first attempt at both the free * and cache lists we try for the correct color. */ /* * XXXX-how do we deal with virtual indexed * caches and and colors? */ VM_STAT_ADD(page_create_cnt[4]); /* * Get lgroup to allocate next page of shared memory * from and use it to specify where to allocate * the physical memory */ lgrp = lgrp_mem_choose(seg, vaddr, PAGESIZE); npp = page_get_freelist(vp, off, seg, vaddr, PAGESIZE, flags | PG_MATCH_COLOR, lgrp); if (npp == NULL) { npp = page_get_cachelist(vp, off, seg, vaddr, flags | PG_MATCH_COLOR, lgrp); if (npp == NULL) { npp = page_create_get_something(vp, off, seg, vaddr, flags & ~PG_MATCH_COLOR); } if (PP_ISAGED(npp) == 0) { /* * Since this page came from the * cachelist, we must destroy the * old vnode association. */ page_hashout(npp, NULL); } } } /* * We own this page! */ ASSERT(PAGE_EXCL(npp)); ASSERT(npp->p_vnode == NULL); ASSERT(!hat_page_is_mapped(npp)); PP_CLRFREE(npp); PP_CLRAGED(npp); /* * Here we have a page in our hot little mits and are * just waiting to stuff it on the appropriate lists. * Get the mutex and check to see if it really does * not exist. */ phm = PAGE_HASH_MUTEX(index); mutex_enter(phm); pp = page_hash_search(index, vp, off); if (pp == NULL) { VM_STAT_ADD(page_create_new); pp = npp; npp = NULL; if (!page_hashin(pp, vp, off, phm)) { /* * Since we hold the page hash mutex and * just searched for this page, page_hashin * had better not fail. If it does, that * means somethread did not follow the * page hash mutex rules. Panic now and * get it over with. As usual, go down * holding all the locks. */ ASSERT(MUTEX_HELD(phm)); panic("page_create: " "hashin failed %p %p %llx %p", (void *)pp, (void *)vp, off, (void *)phm); /*NOTREACHED*/ } ASSERT(MUTEX_HELD(phm)); mutex_exit(phm); phm = NULL; /* * Hat layer locking need not be done to set * the following bits since the page is not hashed * and was on the free list (i.e., had no mappings). * * Set the reference bit to protect * against immediate pageout * * XXXmh modify freelist code to set reference * bit so we don't have to do it here. */ page_set_props(pp, P_REF); found_on_free++; } else { VM_STAT_ADD(page_create_exists); if (flags & PG_EXCL) { /* * Found an existing page, and the caller * wanted all new pages. Undo all of the work * we have done. */ mutex_exit(phm); phm = NULL; while (plist != NULL) { pp = plist; page_sub(&plist, pp); page_io_unlock(pp); /* large pages should not end up here */ ASSERT(pp->p_szc == 0); /*LINTED: constant in conditional ctx*/ VN_DISPOSE(pp, B_INVAL, 0, kcred); } VM_STAT_ADD(page_create_found_one); goto fail; } ASSERT(flags & PG_WAIT); if (!page_lock(pp, SE_EXCL, phm, P_NO_RECLAIM)) { /* * Start all over again if we blocked trying * to lock the page. */ mutex_exit(phm); VM_STAT_ADD(page_create_page_lock_failed); phm = NULL; goto top; } mutex_exit(phm); phm = NULL; if (PP_ISFREE(pp)) { ASSERT(PP_ISAGED(pp) == 0); VM_STAT_ADD(pagecnt.pc_get_cache); page_list_sub(pp, PG_CACHE_LIST); PP_CLRFREE(pp); found_on_free++; } } /* * Got a page! It is locked. Acquire the i/o * lock since we are going to use the p_next and * p_prev fields to link the requested pages together. */ page_io_lock(pp); page_add(&plist, pp); plist = plist->p_next; off += PAGESIZE; vaddr += PAGESIZE; } ASSERT((flags & PG_EXCL) ? (found_on_free == pages_req) : 1); fail: if (npp != NULL) { /* * Did not need this page after all. * Put it back on the free list. */ VM_STAT_ADD(page_create_putbacks); PP_SETFREE(npp); PP_SETAGED(npp); npp->p_offset = (u_offset_t)-1; page_list_add(npp, PG_FREE_LIST | PG_LIST_TAIL); page_unlock(npp); } ASSERT(pages_req >= found_on_free); { uint_t overshoot = (uint_t)(pages_req - found_on_free); if (overshoot) { VM_STAT_ADD(page_create_overshoot); p = &pcf[PCF_INDEX()]; mutex_enter(&p->pcf_lock); if (p->pcf_block) { p->pcf_reserve += overshoot; } else { p->pcf_count += overshoot; if (p->pcf_wait) { mutex_enter(&new_freemem_lock); if (freemem_wait) { cv_signal(&freemem_cv); p->pcf_wait--; } else { p->pcf_wait = 0; } mutex_exit(&new_freemem_lock); } } mutex_exit(&p->pcf_lock); /* freemem is approximate, so this test OK */ if (!p->pcf_block) freemem += overshoot; } } return (plist); } /* * One or more constituent pages of this large page has been marked * toxic. Simply demote the large page to PAGESIZE pages and let * page_free() handle it. This routine should only be called by * large page free routines (page_free_pages() and page_destroy_pages(). * All pages are locked SE_EXCL and have already been marked free. */ static void page_free_toxic_pages(page_t *rootpp) { page_t *tpp; pgcnt_t i, pgcnt = page_get_pagecnt(rootpp->p_szc); uint_t szc = rootpp->p_szc; for (i = 0, tpp = rootpp; i < pgcnt; i++, tpp = tpp->p_next) { ASSERT(tpp->p_szc == szc); ASSERT((PAGE_EXCL(tpp) && !page_iolock_assert(tpp)) || panicstr); tpp->p_szc = 0; } while (rootpp != NULL) { tpp = rootpp; page_sub(&rootpp, tpp); ASSERT(PP_ISFREE(tpp)); PP_CLRFREE(tpp); page_free(tpp, 1); } } /* * Put page on the "free" list. * The free list is really two lists maintained by * the PSM of whatever machine we happen to be on. */ void page_free(page_t *pp, int dontneed) { struct pcf *p; uint_t pcf_index; ASSERT((PAGE_EXCL(pp) && !page_iolock_assert(pp)) || panicstr); if (PP_ISFREE(pp)) { panic("page_free: page %p is free", (void *)pp); } if (pp->p_szc != 0) { if (pp->p_vnode == NULL || IS_SWAPFSVP(pp->p_vnode) || PP_ISKAS(pp)) { panic("page_free: anon or kernel " "or no vnode large page %p", (void *)pp); } page_demote_vp_pages(pp); ASSERT(pp->p_szc == 0); } /* * The page_struct_lock need not be acquired to examine these * fields since the page has an "exclusive" lock. */ if (hat_page_is_mapped(pp) || pp->p_lckcnt != 0 || pp->p_cowcnt != 0 || pp->p_slckcnt != 0) { panic("page_free pp=%p, pfn=%lx, lckcnt=%d, cowcnt=%d " "slckcnt = %d", (void *)pp, page_pptonum(pp), pp->p_lckcnt, pp->p_cowcnt, pp->p_slckcnt); /*NOTREACHED*/ } ASSERT(!hat_page_getshare(pp)); PP_SETFREE(pp); ASSERT(pp->p_vnode == NULL || !IS_VMODSORT(pp->p_vnode) || !hat_ismod(pp)); page_clr_all_props(pp); ASSERT(!hat_page_getshare(pp)); /* * Now we add the page to the head of the free list. * But if this page is associated with a paged vnode * then we adjust the head forward so that the page is * effectively at the end of the list. */ if (pp->p_vnode == NULL) { /* * Page has no identity, put it on the free list. */ PP_SETAGED(pp); pp->p_offset = (u_offset_t)-1; page_list_add(pp, PG_FREE_LIST | PG_LIST_TAIL); VM_STAT_ADD(pagecnt.pc_free_free); TRACE_1(TR_FAC_VM, TR_PAGE_FREE_FREE, "page_free_free:pp %p", pp); } else { PP_CLRAGED(pp); if (!dontneed) { /* move it to the tail of the list */ page_list_add(pp, PG_CACHE_LIST | PG_LIST_TAIL); VM_STAT_ADD(pagecnt.pc_free_cache); TRACE_1(TR_FAC_VM, TR_PAGE_FREE_CACHE_TAIL, "page_free_cache_tail:pp %p", pp); } else { page_list_add(pp, PG_CACHE_LIST | PG_LIST_HEAD); VM_STAT_ADD(pagecnt.pc_free_dontneed); TRACE_1(TR_FAC_VM, TR_PAGE_FREE_CACHE_HEAD, "page_free_cache_head:pp %p", pp); } } page_unlock(pp); /* * Now do the `freemem' accounting. */ pcf_index = PCF_INDEX(); p = &pcf[pcf_index]; mutex_enter(&p->pcf_lock); if (p->pcf_block) { p->pcf_reserve += 1; } else { p->pcf_count += 1; if (p->pcf_wait) { mutex_enter(&new_freemem_lock); /* * Check to see if some other thread * is actually waiting. Another bucket * may have woken it up by now. If there * are no waiters, then set our pcf_wait * count to zero to avoid coming in here * next time. Also, since only one page * was put on the free list, just wake * up one waiter. */ if (freemem_wait) { cv_signal(&freemem_cv); p->pcf_wait--; } else { p->pcf_wait = 0; } mutex_exit(&new_freemem_lock); } } mutex_exit(&p->pcf_lock); /* freemem is approximate, so this test OK */ if (!p->pcf_block) freemem += 1; } /* * Put page on the "free" list during intial startup. * This happens during initial single threaded execution. */ void page_free_at_startup(page_t *pp) { struct pcf *p; uint_t pcf_index; page_list_add(pp, PG_FREE_LIST | PG_LIST_HEAD | PG_LIST_ISINIT); VM_STAT_ADD(pagecnt.pc_free_free); /* * Now do the `freemem' accounting. */ pcf_index = PCF_INDEX(); p = &pcf[pcf_index]; ASSERT(p->pcf_block == 0); ASSERT(p->pcf_wait == 0); p->pcf_count += 1; /* freemem is approximate, so this is OK */ freemem += 1; } void page_free_pages(page_t *pp) { page_t *tpp, *rootpp = NULL; pgcnt_t pgcnt = page_get_pagecnt(pp->p_szc); pgcnt_t i; uint_t szc = pp->p_szc; VM_STAT_ADD(pagecnt.pc_free_pages); TRACE_1(TR_FAC_VM, TR_PAGE_FREE_FREE, "page_free_free:pp %p", pp); ASSERT(pp->p_szc != 0 && pp->p_szc < page_num_pagesizes()); if ((page_pptonum(pp) & (pgcnt - 1)) != 0) { panic("page_free_pages: not root page %p", (void *)pp); /*NOTREACHED*/ } for (i = 0, tpp = pp; i < pgcnt; i++, tpp++) { ASSERT((PAGE_EXCL(tpp) && !page_iolock_assert(tpp)) || panicstr); if (PP_ISFREE(tpp)) { panic("page_free_pages: page %p is free", (void *)tpp); /*NOTREACHED*/ } if (hat_page_is_mapped(tpp) || tpp->p_lckcnt != 0 || tpp->p_cowcnt != 0 || tpp->p_slckcnt != 0) { panic("page_free_pages %p", (void *)tpp); /*NOTREACHED*/ } ASSERT(!hat_page_getshare(tpp)); ASSERT(tpp->p_vnode == NULL); ASSERT(tpp->p_szc == szc); PP_SETFREE(tpp); page_clr_all_props(tpp); PP_SETAGED(tpp); tpp->p_offset = (u_offset_t)-1; ASSERT(tpp->p_next == tpp); ASSERT(tpp->p_prev == tpp); page_list_concat(&rootpp, &tpp); } ASSERT(rootpp == pp); page_list_add_pages(rootpp, 0); page_create_putback(pgcnt); } int free_pages = 1; /* * This routine attempts to return pages to the cachelist via page_release(). * It does not *have* to be successful in all cases, since the pageout scanner * will catch any pages it misses. It does need to be fast and not introduce * too much overhead. * * If a page isn't found on the unlocked sweep of the page_hash bucket, we * don't lock and retry. This is ok, since the page scanner will eventually * find any page we miss in free_vp_pages(). */ void free_vp_pages(vnode_t *vp, u_offset_t off, size_t len) { page_t *pp; u_offset_t eoff; extern int swap_in_range(vnode_t *, u_offset_t, size_t); eoff = off + len; if (free_pages == 0) return; if (swap_in_range(vp, off, len)) return; for (; off < eoff; off += PAGESIZE) { /* * find the page using a fast, but inexact search. It'll be OK * if a few pages slip through the cracks here. */ pp = page_exists(vp, off); /* * If we didn't find the page (it may not exist), the page * is free, looks still in use (shared), or we can't lock it, * just give up. */ if (pp == NULL || PP_ISFREE(pp) || page_share_cnt(pp) > 0 || !page_trylock(pp, SE_EXCL)) continue; /* * Once we have locked pp, verify that it's still the * correct page and not already free */ ASSERT(PAGE_LOCKED_SE(pp, SE_EXCL)); if (pp->p_vnode != vp || pp->p_offset != off || PP_ISFREE(pp)) { page_unlock(pp); continue; } /* * try to release the page... */ (void) page_release(pp, 1); } } /* * Reclaim the given page from the free list. * If pp is part of a large pages, only the given constituent page is reclaimed * and the large page it belonged to will be demoted. This can only happen * if the page is not on the cachelist. * * Returns 1 on success or 0 on failure. * * The page is unlocked if it can't be reclaimed (when freemem == 0). * If `lock' is non-null, it will be dropped and re-acquired if * the routine must wait while freemem is 0. * * As it turns out, boot_getpages() does this. It picks a page, * based on where OBP mapped in some address, gets its pfn, searches * the memsegs, locks the page, then pulls it off the free list! */ int page_reclaim(page_t *pp, kmutex_t *lock) { struct pcf *p; struct cpu *cpup; int enough; uint_t i; ASSERT(lock != NULL ? MUTEX_HELD(lock) : 1); ASSERT(PAGE_EXCL(pp) && PP_ISFREE(pp)); /* * If `freemem' is 0, we cannot reclaim this page from the * freelist, so release every lock we might hold: the page, * and the `lock' before blocking. * * The only way `freemem' can become 0 while there are pages * marked free (have their p->p_free bit set) is when the * system is low on memory and doing a page_create(). In * order to guarantee that once page_create() starts acquiring * pages it will be able to get all that it needs since `freemem' * was decreased by the requested amount. So, we need to release * this page, and let page_create() have it. * * Since `freemem' being zero is not supposed to happen, just * use the usual hash stuff as a starting point. If that bucket * is empty, then assume the worst, and start at the beginning * of the pcf array. If we always start at the beginning * when acquiring more than one pcf lock, there won't be any * deadlock problems. */ /* TODO: Do we need to test kcage_freemem if PG_NORELOC(pp)? */ if (freemem <= throttlefree && !page_create_throttle(1l, 0)) { pcf_acquire_all(); goto page_reclaim_nomem; } enough = pcf_decrement_bucket(1); if (!enough) { VM_STAT_ADD(page_reclaim_zero); /* * Check again. Its possible that some other thread * could have been right behind us, and added one * to a list somewhere. Acquire each of the pcf locks * until we find a page. */ p = pcf; for (i = 0; i < pcf_fanout; i++) { mutex_enter(&p->pcf_lock); if (p->pcf_count >= 1) { p->pcf_count -= 1; /* * freemem is not protected by any lock. Thus, * we cannot have any assertion containing * freemem here. */ freemem -= 1; enough = 1; break; } p++; } if (!enough) { page_reclaim_nomem: /* * We really can't have page `pp'. * Time for the no-memory dance with * page_free(). This is just like * page_create_wait(). Plus the added * attraction of releasing whatever mutex * we held when we were called with in `lock'. * Page_unlock() will wakeup any thread * waiting around for this page. */ if (lock) { VM_STAT_ADD(page_reclaim_zero_locked); mutex_exit(lock); } page_unlock(pp); /* * get this before we drop all the pcf locks. */ mutex_enter(&new_freemem_lock); p = pcf; for (i = 0; i < pcf_fanout; i++) { p->pcf_wait++; mutex_exit(&p->pcf_lock); p++; } freemem_wait++; cv_wait(&freemem_cv, &new_freemem_lock); freemem_wait--; mutex_exit(&new_freemem_lock); if (lock) { mutex_enter(lock); } return (0); } /* * The pcf accounting has been done, * though none of the pcf_wait flags have been set, * drop the locks and continue on. */ while (p >= pcf) { mutex_exit(&p->pcf_lock); p--; } } VM_STAT_ADD(pagecnt.pc_reclaim); /* * page_list_sub will handle the case where pp is a large page. * It's possible that the page was promoted while on the freelist */ if (PP_ISAGED(pp)) { page_list_sub(pp, PG_FREE_LIST); TRACE_1(TR_FAC_VM, TR_PAGE_UNFREE_FREE, "page_reclaim_free:pp %p", pp); } else { page_list_sub(pp, PG_CACHE_LIST); TRACE_1(TR_FAC_VM, TR_PAGE_UNFREE_CACHE, "page_reclaim_cache:pp %p", pp); } /* * clear the p_free & p_age bits since this page is no longer * on the free list. Notice that there was a brief time where * a page is marked as free, but is not on the list. * * Set the reference bit to protect against immediate pageout. */ PP_CLRFREE(pp); PP_CLRAGED(pp); page_set_props(pp, P_REF); CPU_STATS_ENTER_K(); cpup = CPU; /* get cpup now that CPU cannot change */ CPU_STATS_ADDQ(cpup, vm, pgrec, 1); CPU_STATS_ADDQ(cpup, vm, pgfrec, 1); CPU_STATS_EXIT_K(); ASSERT(pp->p_szc == 0); return (1); } /* * Destroy identity of the page and put it back on * the page free list. Assumes that the caller has * acquired the "exclusive" lock on the page. */ void page_destroy(page_t *pp, int dontfree) { ASSERT((PAGE_EXCL(pp) && !page_iolock_assert(pp)) || panicstr); ASSERT(pp->p_slckcnt == 0 || panicstr); if (pp->p_szc != 0) { if (pp->p_vnode == NULL || IS_SWAPFSVP(pp->p_vnode) || PP_ISKAS(pp)) { panic("page_destroy: anon or kernel or no vnode " "large page %p", (void *)pp); } page_demote_vp_pages(pp); ASSERT(pp->p_szc == 0); } TRACE_1(TR_FAC_VM, TR_PAGE_DESTROY, "page_destroy:pp %p", pp); /* * Unload translations, if any, then hash out the * page to erase its identity. */ (void) hat_pageunload(pp, HAT_FORCE_PGUNLOAD); page_hashout(pp, NULL); if (!dontfree) { /* * Acquire the "freemem_lock" for availrmem. * The page_struct_lock need not be acquired for lckcnt * and cowcnt since the page has an "exclusive" lock. * We are doing a modified version of page_pp_unlock here. */ if ((pp->p_lckcnt != 0) || (pp->p_cowcnt != 0)) { mutex_enter(&freemem_lock); if (pp->p_lckcnt != 0) { availrmem++; pages_locked--; pp->p_lckcnt = 0; } if (pp->p_cowcnt != 0) { availrmem += pp->p_cowcnt; pages_locked -= pp->p_cowcnt; pp->p_cowcnt = 0; } mutex_exit(&freemem_lock); } /* * Put the page on the "free" list. */ page_free(pp, 0); } } void page_destroy_pages(page_t *pp) { page_t *tpp, *rootpp = NULL; pgcnt_t pgcnt = page_get_pagecnt(pp->p_szc); pgcnt_t i, pglcks = 0; uint_t szc = pp->p_szc; ASSERT(pp->p_szc != 0 && pp->p_szc < page_num_pagesizes()); VM_STAT_ADD(pagecnt.pc_destroy_pages); TRACE_1(TR_FAC_VM, TR_PAGE_DESTROY, "page_destroy_pages:pp %p", pp); if ((page_pptonum(pp) & (pgcnt - 1)) != 0) { panic("page_destroy_pages: not root page %p", (void *)pp); /*NOTREACHED*/ } for (i = 0, tpp = pp; i < pgcnt; i++, tpp++) { ASSERT((PAGE_EXCL(tpp) && !page_iolock_assert(tpp)) || panicstr); ASSERT(tpp->p_slckcnt == 0 || panicstr); (void) hat_pageunload(tpp, HAT_FORCE_PGUNLOAD); page_hashout(tpp, NULL); ASSERT(tpp->p_offset == (u_offset_t)-1); if (tpp->p_lckcnt != 0) { pglcks++; tpp->p_lckcnt = 0; } else if (tpp->p_cowcnt != 0) { pglcks += tpp->p_cowcnt; tpp->p_cowcnt = 0; } ASSERT(!hat_page_getshare(tpp)); ASSERT(tpp->p_vnode == NULL); ASSERT(tpp->p_szc == szc); PP_SETFREE(tpp); page_clr_all_props(tpp); PP_SETAGED(tpp); ASSERT(tpp->p_next == tpp); ASSERT(tpp->p_prev == tpp); page_list_concat(&rootpp, &tpp); } ASSERT(rootpp == pp); if (pglcks != 0) { mutex_enter(&freemem_lock); availrmem += pglcks; mutex_exit(&freemem_lock); } page_list_add_pages(rootpp, 0); page_create_putback(pgcnt); } /* * Similar to page_destroy(), but destroys pages which are * locked and known to be on the page free list. Since * the page is known to be free and locked, no one can access * it. * * Also, the number of free pages does not change. */ void page_destroy_free(page_t *pp) { ASSERT(PAGE_EXCL(pp)); ASSERT(PP_ISFREE(pp)); ASSERT(pp->p_vnode); ASSERT(hat_page_getattr(pp, P_MOD | P_REF | P_RO) == 0); ASSERT(!hat_page_is_mapped(pp)); ASSERT(PP_ISAGED(pp) == 0); ASSERT(pp->p_szc == 0); VM_STAT_ADD(pagecnt.pc_destroy_free); page_list_sub(pp, PG_CACHE_LIST); page_hashout(pp, NULL); ASSERT(pp->p_vnode == NULL); ASSERT(pp->p_offset == (u_offset_t)-1); ASSERT(pp->p_hash == NULL); PP_SETAGED(pp); page_list_add(pp, PG_FREE_LIST | PG_LIST_TAIL); page_unlock(pp); mutex_enter(&new_freemem_lock); if (freemem_wait) { cv_signal(&freemem_cv); } mutex_exit(&new_freemem_lock); } /* * Rename the page "opp" to have an identity specified * by [vp, off]. If a page already exists with this name * it is locked and destroyed. Note that the page's * translations are not unloaded during the rename. * * This routine is used by the anon layer to "steal" the * original page and is not unlike destroying a page and * creating a new page using the same page frame. * * XXX -- Could deadlock if caller 1 tries to rename A to B while * caller 2 tries to rename B to A. */ void page_rename(page_t *opp, vnode_t *vp, u_offset_t off) { page_t *pp; int olckcnt = 0; int ocowcnt = 0; kmutex_t *phm; ulong_t index; ASSERT(PAGE_EXCL(opp) && !page_iolock_assert(opp)); ASSERT(MUTEX_NOT_HELD(page_vnode_mutex(vp))); ASSERT(PP_ISFREE(opp) == 0); VM_STAT_ADD(page_rename_count); TRACE_3(TR_FAC_VM, TR_PAGE_RENAME, "page rename:pp %p vp %p off %llx", opp, vp, off); /* * CacheFS may call page_rename for a large NFS page * when both CacheFS and NFS mount points are used * by applications. Demote this large page before * renaming it, to ensure that there are no "partial" * large pages left lying around. */ if (opp->p_szc != 0) { vnode_t *ovp = opp->p_vnode; ASSERT(ovp != NULL); ASSERT(!IS_SWAPFSVP(ovp)); ASSERT(!VN_ISKAS(ovp)); page_demote_vp_pages(opp); ASSERT(opp->p_szc == 0); } page_hashout(opp, NULL); PP_CLRAGED(opp); /* * Acquire the appropriate page hash lock, since * we're going to rename the page. */ index = PAGE_HASH_FUNC(vp, off); phm = PAGE_HASH_MUTEX(index); mutex_enter(phm); top: /* * Look for an existing page with this name and destroy it if found. * By holding the page hash lock all the way to the page_hashin() * call, we are assured that no page can be created with this * identity. In the case when the phm lock is dropped to undo any * hat layer mappings, the existing page is held with an "exclusive" * lock, again preventing another page from being created with * this identity. */ pp = page_hash_search(index, vp, off); if (pp != NULL) { VM_STAT_ADD(page_rename_exists); /* * As it turns out, this is one of only two places where * page_lock() needs to hold the passed in lock in the * successful case. In all of the others, the lock could * be dropped as soon as the attempt is made to lock * the page. It is tempting to add yet another arguement, * PL_KEEP or PL_DROP, to let page_lock know what to do. */ if (!page_lock(pp, SE_EXCL, phm, P_RECLAIM)) { /* * Went to sleep because the page could not * be locked. We were woken up when the page * was unlocked, or when the page was destroyed. * In either case, `phm' was dropped while we * slept. Hence we should not just roar through * this loop. */ goto top; } /* * If an existing page is a large page, then demote * it to ensure that no "partial" large pages are * "created" after page_rename. An existing page * can be a CacheFS page, and can't belong to swapfs. */ if (hat_page_is_mapped(pp)) { /* * Unload translations. Since we hold the * exclusive lock on this page, the page * can not be changed while we drop phm. * This is also not a lock protocol violation, * but rather the proper way to do things. */ mutex_exit(phm); (void) hat_pageunload(pp, HAT_FORCE_PGUNLOAD); if (pp->p_szc != 0) { ASSERT(!IS_SWAPFSVP(vp)); ASSERT(!VN_ISKAS(vp)); page_demote_vp_pages(pp); ASSERT(pp->p_szc == 0); } mutex_enter(phm); } else if (pp->p_szc != 0) { ASSERT(!IS_SWAPFSVP(vp)); ASSERT(!VN_ISKAS(vp)); mutex_exit(phm); page_demote_vp_pages(pp); ASSERT(pp->p_szc == 0); mutex_enter(phm); } page_hashout(pp, phm); } /* * Hash in the page with the new identity. */ if (!page_hashin(opp, vp, off, phm)) { /* * We were holding phm while we searched for [vp, off] * and only dropped phm if we found and locked a page. * If we can't create this page now, then some thing * is really broken. */ panic("page_rename: Can't hash in page: %p", (void *)pp); /*NOTREACHED*/ } ASSERT(MUTEX_HELD(phm)); mutex_exit(phm); /* * Now that we have dropped phm, lets get around to finishing up * with pp. */ if (pp != NULL) { ASSERT(!hat_page_is_mapped(pp)); /* for now large pages should not end up here */ ASSERT(pp->p_szc == 0); /* * Save the locks for transfer to the new page and then * clear them so page_free doesn't think they're important. * The page_struct_lock need not be acquired for lckcnt and * cowcnt since the page has an "exclusive" lock. */ olckcnt = pp->p_lckcnt; ocowcnt = pp->p_cowcnt; pp->p_lckcnt = pp->p_cowcnt = 0; /* * Put the page on the "free" list after we drop * the lock. The less work under the lock the better. */ /*LINTED: constant in conditional context*/ VN_DISPOSE(pp, B_FREE, 0, kcred); } /* * Transfer the lock count from the old page (if any). * The page_struct_lock need not be acquired for lckcnt and * cowcnt since the page has an "exclusive" lock. */ opp->p_lckcnt += olckcnt; opp->p_cowcnt += ocowcnt; } /* * low level routine to add page `pp' to the hash and vp chains for [vp, offset] * * Pages are normally inserted at the start of a vnode's v_pages list. * If the vnode is VMODSORT and the page is modified, it goes at the end. * This can happen when a modified page is relocated for DR. * * Returns 1 on success and 0 on failure. */ static int page_do_hashin(page_t *pp, vnode_t *vp, u_offset_t offset) { page_t **listp; page_t *tp; ulong_t index; ASSERT(PAGE_EXCL(pp)); ASSERT(vp != NULL); ASSERT(MUTEX_HELD(page_vnode_mutex(vp))); /* * Be sure to set these up before the page is inserted on the hash * list. As soon as the page is placed on the list some other * thread might get confused and wonder how this page could * possibly hash to this list. */ pp->p_vnode = vp; pp->p_offset = offset; /* * record if this page is on a swap vnode */ if ((vp->v_flag & VISSWAP) != 0) PP_SETSWAP(pp); index = PAGE_HASH_FUNC(vp, offset); ASSERT(MUTEX_HELD(PAGE_HASH_MUTEX(index))); listp = &page_hash[index]; /* * If this page is already hashed in, fail this attempt to add it. */ for (tp = *listp; tp != NULL; tp = tp->p_hash) { if (tp->p_vnode == vp && tp->p_offset == offset) { pp->p_vnode = NULL; pp->p_offset = (u_offset_t)(-1); return (0); } } pp->p_hash = *listp; *listp = pp; /* * Add the page to the vnode's list of pages */ if (vp->v_pages != NULL && IS_VMODSORT(vp) && hat_ismod(pp)) listp = &vp->v_pages->p_vpprev->p_vpnext; else listp = &vp->v_pages; page_vpadd(listp, pp); return (1); } /* * Add page `pp' to both the hash and vp chains for [vp, offset]. * * Returns 1 on success and 0 on failure. * If hold is passed in, it is not dropped. */ int page_hashin(page_t *pp, vnode_t *vp, u_offset_t offset, kmutex_t *hold) { kmutex_t *phm = NULL; kmutex_t *vphm; int rc; ASSERT(MUTEX_NOT_HELD(page_vnode_mutex(vp))); ASSERT(pp->p_fsdata == 0 || panicstr); TRACE_3(TR_FAC_VM, TR_PAGE_HASHIN, "page_hashin:pp %p vp %p offset %llx", pp, vp, offset); VM_STAT_ADD(hashin_count); if (hold != NULL) phm = hold; else { VM_STAT_ADD(hashin_not_held); phm = PAGE_HASH_MUTEX(PAGE_HASH_FUNC(vp, offset)); mutex_enter(phm); } vphm = page_vnode_mutex(vp); mutex_enter(vphm); rc = page_do_hashin(pp, vp, offset); mutex_exit(vphm); if (hold == NULL) mutex_exit(phm); if (rc == 0) VM_STAT_ADD(hashin_already); return (rc); } /* * Remove page ``pp'' from the hash and vp chains and remove vp association. * All mutexes must be held */ static void page_do_hashout(page_t *pp) { page_t **hpp; page_t *hp; vnode_t *vp = pp->p_vnode; ASSERT(vp != NULL); ASSERT(MUTEX_HELD(page_vnode_mutex(vp))); /* * First, take pp off of its hash chain. */ hpp = &page_hash[PAGE_HASH_FUNC(vp, pp->p_offset)]; for (;;) { hp = *hpp; if (hp == pp) break; if (hp == NULL) { panic("page_do_hashout"); /*NOTREACHED*/ } hpp = &hp->p_hash; } *hpp = pp->p_hash; /* * Now remove it from its associated vnode. */ if (vp->v_pages) page_vpsub(&vp->v_pages, pp); pp->p_hash = NULL; page_clr_all_props(pp); PP_CLRSWAP(pp); pp->p_vnode = NULL; pp->p_offset = (u_offset_t)-1; pp->p_fsdata = 0; } /* * Remove page ``pp'' from the hash and vp chains and remove vp association. * * When `phm' is non-NULL it contains the address of the mutex protecting the * hash list pp is on. It is not dropped. */ void page_hashout(page_t *pp, kmutex_t *phm) { vnode_t *vp; ulong_t index; kmutex_t *nphm; kmutex_t *vphm; kmutex_t *sep; ASSERT(phm != NULL ? MUTEX_HELD(phm) : 1); ASSERT(pp->p_vnode != NULL); ASSERT((PAGE_EXCL(pp) && !page_iolock_assert(pp)) || panicstr); ASSERT(MUTEX_NOT_HELD(page_vnode_mutex(pp->p_vnode))); vp = pp->p_vnode; TRACE_2(TR_FAC_VM, TR_PAGE_HASHOUT, "page_hashout:pp %p vp %p", pp, vp); /* * */ VM_STAT_ADD(hashout_count); index = PAGE_HASH_FUNC(vp, pp->p_offset); if (phm == NULL) { VM_STAT_ADD(hashout_not_held); nphm = PAGE_HASH_MUTEX(index); mutex_enter(nphm); } ASSERT(phm ? phm == PAGE_HASH_MUTEX(index) : 1); /* * grab page vnode mutex and remove it... */ vphm = page_vnode_mutex(vp); mutex_enter(vphm); page_do_hashout(pp); mutex_exit(vphm); if (phm == NULL) mutex_exit(nphm); /* * Wake up processes waiting for this page. The page's * identity has been changed, and is probably not the * desired page any longer. */ sep = page_se_mutex(pp); mutex_enter(sep); pp->p_selock &= ~SE_EWANTED; if (CV_HAS_WAITERS(&pp->p_cv)) cv_broadcast(&pp->p_cv); mutex_exit(sep); } /* * Add the page to the front of a linked list of pages * using the p_next & p_prev pointers for the list. * The caller is responsible for protecting the list pointers. */ void page_add(page_t **ppp, page_t *pp) { ASSERT(PAGE_EXCL(pp) || (PAGE_SHARED(pp) && page_iolock_assert(pp))); page_add_common(ppp, pp); } /* * Common code for page_add() and mach_page_add() */ void page_add_common(page_t **ppp, page_t *pp) { if (*ppp == NULL) { pp->p_next = pp->p_prev = pp; } else { pp->p_next = *ppp; pp->p_prev = (*ppp)->p_prev; (*ppp)->p_prev = pp; pp->p_prev->p_next = pp; } *ppp = pp; } /* * Remove this page from a linked list of pages * using the p_next & p_prev pointers for the list. * * The caller is responsible for protecting the list pointers. */ void page_sub(page_t **ppp, page_t *pp) { ASSERT(pp != NULL && (PP_ISFREE(pp)) ? 1 : (PAGE_EXCL(pp)) || (PAGE_SHARED(pp) && page_iolock_assert(pp))); if (*ppp == NULL || pp == NULL) { panic("page_sub: bad arg(s): pp %p, *ppp %p", (void *)pp, (void *)(*ppp)); /*NOTREACHED*/ } page_sub_common(ppp, pp); } /* * Common code for page_sub() and mach_page_sub() */ void page_sub_common(page_t **ppp, page_t *pp) { if (*ppp == pp) *ppp = pp->p_next; /* go to next page */ if (*ppp == pp) *ppp = NULL; /* page list is gone */ else { pp->p_prev->p_next = pp->p_next; pp->p_next->p_prev = pp->p_prev; } pp->p_prev = pp->p_next = pp; /* make pp a list of one */ } /* * Break page list cppp into two lists with npages in the first list. * The tail is returned in nppp. */ void page_list_break(page_t **oppp, page_t **nppp, pgcnt_t npages) { page_t *s1pp = *oppp; page_t *s2pp; page_t *e1pp, *e2pp; long n = 0; if (s1pp == NULL) { *nppp = NULL; return; } if (npages == 0) { *nppp = s1pp; *oppp = NULL; return; } for (n = 0, s2pp = *oppp; n < npages; n++) { s2pp = s2pp->p_next; } /* Fix head and tail of new lists */ e1pp = s2pp->p_prev; e2pp = s1pp->p_prev; s1pp->p_prev = e1pp; e1pp->p_next = s1pp; s2pp->p_prev = e2pp; e2pp->p_next = s2pp; /* second list empty */ if (s2pp == s1pp) { *oppp = s1pp; *nppp = NULL; } else { *oppp = s1pp; *nppp = s2pp; } } /* * Concatenate page list nppp onto the end of list ppp. */ void page_list_concat(page_t **ppp, page_t **nppp) { page_t *s1pp, *s2pp, *e1pp, *e2pp; if (*nppp == NULL) { return; } if (*ppp == NULL) { *ppp = *nppp; return; } s1pp = *ppp; e1pp = s1pp->p_prev; s2pp = *nppp; e2pp = s2pp->p_prev; s1pp->p_prev = e2pp; e2pp->p_next = s1pp; e1pp->p_next = s2pp; s2pp->p_prev = e1pp; } /* * return the next page in the page list */ page_t * page_list_next(page_t *pp) { return (pp->p_next); } /* * Add the page to the front of the linked list of pages * using p_vpnext/p_vpprev pointers for the list. * * The caller is responsible for protecting the lists. */ void page_vpadd(page_t **ppp, page_t *pp) { if (*ppp == NULL) { pp->p_vpnext = pp->p_vpprev = pp; } else { pp->p_vpnext = *ppp; pp->p_vpprev = (*ppp)->p_vpprev; (*ppp)->p_vpprev = pp; pp->p_vpprev->p_vpnext = pp; } *ppp = pp; } /* * Remove this page from the linked list of pages * using p_vpnext/p_vpprev pointers for the list. * * The caller is responsible for protecting the lists. */ void page_vpsub(page_t **ppp, page_t *pp) { if (*ppp == NULL || pp == NULL) { panic("page_vpsub: bad arg(s): pp %p, *ppp %p", (void *)pp, (void *)(*ppp)); /*NOTREACHED*/ } if (*ppp == pp) *ppp = pp->p_vpnext; /* go to next page */ if (*ppp == pp) *ppp = NULL; /* page list is gone */ else { pp->p_vpprev->p_vpnext = pp->p_vpnext; pp->p_vpnext->p_vpprev = pp->p_vpprev; } pp->p_vpprev = pp->p_vpnext = pp; /* make pp a list of one */ } /* * Lock a physical page into memory "long term". Used to support "lock * in memory" functions. Accepts the page to be locked, and a cow variable * to indicate whether a the lock will travel to the new page during * a potential copy-on-write. */ int page_pp_lock( page_t *pp, /* page to be locked */ int cow, /* cow lock */ int kernel) /* must succeed -- ignore checking */ { int r = 0; /* result -- assume failure */ ASSERT(PAGE_LOCKED(pp)); page_struct_lock(pp); /* * Acquire the "freemem_lock" for availrmem. */ if (cow) { mutex_enter(&freemem_lock); if ((availrmem > pages_pp_maximum) && (pp->p_cowcnt < (ushort_t)PAGE_LOCK_MAXIMUM)) { availrmem--; pages_locked++; mutex_exit(&freemem_lock); r = 1; if (++pp->p_cowcnt == (ushort_t)PAGE_LOCK_MAXIMUM) { cmn_err(CE_WARN, "COW lock limit reached on pfn 0x%lx", page_pptonum(pp)); } } else mutex_exit(&freemem_lock); } else { if (pp->p_lckcnt) { if (pp->p_lckcnt < (ushort_t)PAGE_LOCK_MAXIMUM) { r = 1; if (++pp->p_lckcnt == (ushort_t)PAGE_LOCK_MAXIMUM) { cmn_err(CE_WARN, "Page lock limit " "reached on pfn 0x%lx", page_pptonum(pp)); } } } else { if (kernel) { /* availrmem accounting done by caller */ ++pp->p_lckcnt; r = 1; } else { mutex_enter(&freemem_lock); if (availrmem > pages_pp_maximum) { availrmem--; pages_locked++; ++pp->p_lckcnt; r = 1; } mutex_exit(&freemem_lock); } } } page_struct_unlock(pp); return (r); } /* * Decommit a lock on a physical page frame. Account for cow locks if * appropriate. */ void page_pp_unlock( page_t *pp, /* page to be unlocked */ int cow, /* expect cow lock */ int kernel) /* this was a kernel lock */ { ASSERT(PAGE_LOCKED(pp)); page_struct_lock(pp); /* * Acquire the "freemem_lock" for availrmem. * If cowcnt or lcknt is already 0 do nothing; i.e., we * could be called to unlock even if nothing is locked. This could * happen if locked file pages were truncated (removing the lock) * and the file was grown again and new pages faulted in; the new * pages are unlocked but the segment still thinks they're locked. */ if (cow) { if (pp->p_cowcnt) { mutex_enter(&freemem_lock); pp->p_cowcnt--; availrmem++; pages_locked--; mutex_exit(&freemem_lock); } } else { if (pp->p_lckcnt && --pp->p_lckcnt == 0) { if (!kernel) { mutex_enter(&freemem_lock); availrmem++; pages_locked--; mutex_exit(&freemem_lock); } } } page_struct_unlock(pp); } /* * This routine reserves availrmem for npages. * It returns 1 on success or 0 on failure. * * flags: KM_NOSLEEP or KM_SLEEP * cb_wait: called to induce delay when KM_SLEEP reservation requires kmem * reaping to potentially succeed. If the callback returns 0, the * reservation attempts will cease to repeat and page_xresv() may * report a failure. If cb_wait is NULL, the traditional delay(hz/2) * behavior will be used while waiting for a reap. */ int page_xresv(pgcnt_t npages, uint_t flags, int (*cb_wait)(void)) { mutex_enter(&freemem_lock); if (availrmem >= tune.t_minarmem + npages) { availrmem -= npages; mutex_exit(&freemem_lock); return (1); } else if ((flags & KM_NOSLEEP) != 0) { mutex_exit(&freemem_lock); return (0); } mutex_exit(&freemem_lock); /* * We signal memory pressure to the system by elevating 'needfree'. * Processes such as kmem reaping, pageout, and ZFS ARC shrinking can * then respond to said pressure by freeing pages. */ page_needfree(npages); int nobail = 1; do { kmem_reap(); if (cb_wait == NULL) { delay(hz >> 2); } else { nobail = cb_wait(); } mutex_enter(&freemem_lock); if (availrmem >= tune.t_minarmem + npages) { availrmem -= npages; mutex_exit(&freemem_lock); page_needfree(-(spgcnt_t)npages); return (1); } mutex_exit(&freemem_lock); } while (nobail != 0); page_needfree(-(spgcnt_t)npages); return (0); } /* * This routine reserves availrmem for npages; * flags: KM_NOSLEEP or KM_SLEEP * returns 1 on success or 0 on failure */ int page_resv(pgcnt_t npages, uint_t flags) { return (page_xresv(npages, flags, NULL)); } /* * This routine unreserves availrmem for npages; */ void page_unresv(pgcnt_t npages) { mutex_enter(&freemem_lock); availrmem += npages; mutex_exit(&freemem_lock); } /* * See Statement at the beginning of segvn_lockop() regarding * the way we handle cowcnts and lckcnts. * * Transfer cowcnt on 'opp' to cowcnt on 'npp' if the vpage * that breaks COW has PROT_WRITE. * * Note that, we may also break COW in case we are softlocking * on read access during physio; * in this softlock case, the vpage may not have PROT_WRITE. * So, we need to transfer lckcnt on 'opp' to lckcnt on 'npp' * if the vpage doesn't have PROT_WRITE. * * This routine is never called if we are stealing a page * in anon_private. * * The caller subtracted from availrmem for read only mapping. * if lckcnt is 1 increment availrmem. */ void page_pp_useclaim( page_t *opp, /* original page frame losing lock */ page_t *npp, /* new page frame gaining lock */ uint_t write_perm) /* set if vpage has PROT_WRITE */ { int payback = 0; int nidx, oidx; ASSERT(PAGE_LOCKED(opp)); ASSERT(PAGE_LOCKED(npp)); /* * Since we have two pages we probably have two locks. We need to take * them in a defined order to avoid deadlocks. It's also possible they * both hash to the same lock in which case this is a non-issue. */ nidx = PAGE_LLOCK_HASH(PP_PAGEROOT(npp)); oidx = PAGE_LLOCK_HASH(PP_PAGEROOT(opp)); if (nidx < oidx) { page_struct_lock(npp); page_struct_lock(opp); } else if (oidx < nidx) { page_struct_lock(opp); page_struct_lock(npp); } else { /* The pages hash to the same lock */ page_struct_lock(npp); } ASSERT(npp->p_cowcnt == 0); ASSERT(npp->p_lckcnt == 0); /* Don't use claim if nothing is locked (see page_pp_unlock above) */ if ((write_perm && opp->p_cowcnt != 0) || (!write_perm && opp->p_lckcnt != 0)) { if (write_perm) { npp->p_cowcnt++; ASSERT(opp->p_cowcnt != 0); opp->p_cowcnt--; } else { ASSERT(opp->p_lckcnt != 0); /* * We didn't need availrmem decremented if p_lckcnt on * original page is 1. Here, we are unlocking * read-only copy belonging to original page and * are locking a copy belonging to new page. */ if (opp->p_lckcnt == 1) payback = 1; npp->p_lckcnt++; opp->p_lckcnt--; } } if (payback) { mutex_enter(&freemem_lock); availrmem++; pages_useclaim--; mutex_exit(&freemem_lock); } if (nidx < oidx) { page_struct_unlock(opp); page_struct_unlock(npp); } else if (oidx < nidx) { page_struct_unlock(npp); page_struct_unlock(opp); } else { /* The pages hash to the same lock */ page_struct_unlock(npp); } } /* * Simple claim adjust functions -- used to support changes in * claims due to changes in access permissions. Used by segvn_setprot(). */ int page_addclaim(page_t *pp) { int r = 0; /* result */ ASSERT(PAGE_LOCKED(pp)); page_struct_lock(pp); ASSERT(pp->p_lckcnt != 0); if (pp->p_lckcnt == 1) { if (pp->p_cowcnt < (ushort_t)PAGE_LOCK_MAXIMUM) { --pp->p_lckcnt; r = 1; if (++pp->p_cowcnt == (ushort_t)PAGE_LOCK_MAXIMUM) { cmn_err(CE_WARN, "COW lock limit reached on pfn 0x%lx", page_pptonum(pp)); } } } else { mutex_enter(&freemem_lock); if ((availrmem > pages_pp_maximum) && (pp->p_cowcnt < (ushort_t)PAGE_LOCK_MAXIMUM)) { --availrmem; ++pages_claimed; mutex_exit(&freemem_lock); --pp->p_lckcnt; r = 1; if (++pp->p_cowcnt == (ushort_t)PAGE_LOCK_MAXIMUM) { cmn_err(CE_WARN, "COW lock limit reached on pfn 0x%lx", page_pptonum(pp)); } } else mutex_exit(&freemem_lock); } page_struct_unlock(pp); return (r); } int page_subclaim(page_t *pp) { int r = 0; ASSERT(PAGE_LOCKED(pp)); page_struct_lock(pp); ASSERT(pp->p_cowcnt != 0); if (pp->p_lckcnt) { if (pp->p_lckcnt < (ushort_t)PAGE_LOCK_MAXIMUM) { r = 1; /* * for availrmem */ mutex_enter(&freemem_lock); availrmem++; pages_claimed--; mutex_exit(&freemem_lock); pp->p_cowcnt--; if (++pp->p_lckcnt == (ushort_t)PAGE_LOCK_MAXIMUM) { cmn_err(CE_WARN, "Page lock limit reached on pfn 0x%lx", page_pptonum(pp)); } } } else { r = 1; pp->p_cowcnt--; pp->p_lckcnt++; } page_struct_unlock(pp); return (r); } /* * Variant of page_addclaim(), where ppa[] contains the pages of a single large * page. */ int page_addclaim_pages(page_t **ppa) { pgcnt_t lckpgs = 0, pg_idx; VM_STAT_ADD(pagecnt.pc_addclaim_pages); /* * Only need to take the page struct lock on the large page root. */ page_struct_lock(ppa[0]); for (pg_idx = 0; ppa[pg_idx] != NULL; pg_idx++) { ASSERT(PAGE_LOCKED(ppa[pg_idx])); ASSERT(ppa[pg_idx]->p_lckcnt != 0); if (ppa[pg_idx]->p_cowcnt == (ushort_t)PAGE_LOCK_MAXIMUM) { page_struct_unlock(ppa[0]); return (0); } if (ppa[pg_idx]->p_lckcnt > 1) lckpgs++; } if (lckpgs != 0) { mutex_enter(&freemem_lock); if (availrmem >= pages_pp_maximum + lckpgs) { availrmem -= lckpgs; pages_claimed += lckpgs; } else { mutex_exit(&freemem_lock); page_struct_unlock(ppa[0]); return (0); } mutex_exit(&freemem_lock); } for (pg_idx = 0; ppa[pg_idx] != NULL; pg_idx++) { ppa[pg_idx]->p_lckcnt--; ppa[pg_idx]->p_cowcnt++; } page_struct_unlock(ppa[0]); return (1); } /* * Variant of page_subclaim(), where ppa[] contains the pages of a single large * page. */ int page_subclaim_pages(page_t **ppa) { pgcnt_t ulckpgs = 0, pg_idx; VM_STAT_ADD(pagecnt.pc_subclaim_pages); /* * Only need to take the page struct lock on the large page root. */ page_struct_lock(ppa[0]); for (pg_idx = 0; ppa[pg_idx] != NULL; pg_idx++) { ASSERT(PAGE_LOCKED(ppa[pg_idx])); ASSERT(ppa[pg_idx]->p_cowcnt != 0); if (ppa[pg_idx]->p_lckcnt == (ushort_t)PAGE_LOCK_MAXIMUM) { page_struct_unlock(ppa[0]); return (0); } if (ppa[pg_idx]->p_lckcnt != 0) ulckpgs++; } if (ulckpgs != 0) { mutex_enter(&freemem_lock); availrmem += ulckpgs; pages_claimed -= ulckpgs; mutex_exit(&freemem_lock); } for (pg_idx = 0; ppa[pg_idx] != NULL; pg_idx++) { ppa[pg_idx]->p_cowcnt--; ppa[pg_idx]->p_lckcnt++; } page_struct_unlock(ppa[0]); return (1); } page_t * page_numtopp(pfn_t pfnum, se_t se) { page_t *pp; retry: pp = page_numtopp_nolock(pfnum); if (pp == NULL) { return ((page_t *)NULL); } /* * Acquire the appropriate lock on the page. */ while (!page_lock(pp, se, (kmutex_t *)NULL, P_RECLAIM)) { if (page_pptonum(pp) != pfnum) goto retry; continue; } if (page_pptonum(pp) != pfnum) { page_unlock(pp); goto retry; } return (pp); } page_t * page_numtopp_noreclaim(pfn_t pfnum, se_t se) { page_t *pp; retry: pp = page_numtopp_nolock(pfnum); if (pp == NULL) { return ((page_t *)NULL); } /* * Acquire the appropriate lock on the page. */ while (!page_lock(pp, se, (kmutex_t *)NULL, P_NO_RECLAIM)) { if (page_pptonum(pp) != pfnum) goto retry; continue; } if (page_pptonum(pp) != pfnum) { page_unlock(pp); goto retry; } return (pp); } /* * This routine is like page_numtopp, but will only return page structs * for pages which are ok for loading into hardware using the page struct. */ page_t * page_numtopp_nowait(pfn_t pfnum, se_t se) { page_t *pp; retry: pp = page_numtopp_nolock(pfnum); if (pp == NULL) { return ((page_t *)NULL); } /* * Try to acquire the appropriate lock on the page. */ if (PP_ISFREE(pp)) pp = NULL; else { if (!page_trylock(pp, se)) pp = NULL; else { if (page_pptonum(pp) != pfnum) { page_unlock(pp); goto retry; } if (PP_ISFREE(pp)) { page_unlock(pp); pp = NULL; } } } return (pp); } /* * Returns a count of dirty pages that are in the process * of being written out. If 'cleanit' is set, try to push the page. */ pgcnt_t page_busy(int cleanit) { page_t *page0 = page_first(); page_t *pp = page0; pgcnt_t nppbusy = 0; u_offset_t off; do { vnode_t *vp = pp->p_vnode; /* * A page is a candidate for syncing if it is: * * (a) On neither the freelist nor the cachelist * (b) Hashed onto a vnode * (c) Not a kernel page * (d) Dirty * (e) Not part of a swapfile * (f) a page which belongs to a real vnode; eg has a non-null * v_vfsp pointer. * (g) Backed by a filesystem which doesn't have a * stubbed-out sync operation */ if (!PP_ISFREE(pp) && vp != NULL && !VN_ISKAS(vp) && hat_ismod(pp) && !IS_SWAPVP(vp) && vp->v_vfsp != NULL && vfs_can_sync(vp->v_vfsp)) { nppbusy++; if (!cleanit) continue; if (!page_trylock(pp, SE_EXCL)) continue; if (PP_ISFREE(pp) || vp == NULL || IS_SWAPVP(vp) || pp->p_lckcnt != 0 || pp->p_cowcnt != 0 || !(hat_pagesync(pp, HAT_SYNC_DONTZERO | HAT_SYNC_STOPON_MOD) & P_MOD)) { page_unlock(pp); continue; } off = pp->p_offset; VN_HOLD(vp); page_unlock(pp); (void) VOP_PUTPAGE(vp, off, PAGESIZE, B_ASYNC | B_FREE, kcred, NULL); VN_RELE(vp); } } while ((pp = page_next(pp)) != page0); return (nppbusy); } void page_invalidate_pages(void); /* * callback handler to vm sub-system * * callers make sure no recursive entries to this func. */ /*ARGSUSED*/ boolean_t callb_vm_cpr(void *arg, int code) { if (code == CB_CODE_CPR_CHKPT) page_invalidate_pages(); return (B_TRUE); } /* * Invalidate all pages of the system. * It shouldn't be called until all user page activities are all stopped. */ void page_invalidate_pages() { page_t *pp; page_t *page0; pgcnt_t nbusypages; int retry = 0; const int MAXRETRIES = 4; top: /* * Flush dirty pages and destroy the clean ones. */ nbusypages = 0; pp = page0 = page_first(); do { struct vnode *vp; u_offset_t offset; int mod; /* * skip the page if it has no vnode or the page associated * with the kernel vnode or prom allocated kernel mem. */ if ((vp = pp->p_vnode) == NULL || VN_ISKAS(vp)) continue; /* * skip the page which is already free invalidated. */ if (PP_ISFREE(pp) && PP_ISAGED(pp)) continue; /* * skip pages that are already locked or can't be "exclusively" * locked or are already free. After we lock the page, check * the free and age bits again to be sure it's not destroyed * yet. * To achieve max. parallelization, we use page_trylock instead * of page_lock so that we don't get block on individual pages * while we have thousands of other pages to process. */ if (!page_trylock(pp, SE_EXCL)) { nbusypages++; continue; } else if (PP_ISFREE(pp)) { if (!PP_ISAGED(pp)) { page_destroy_free(pp); } else { page_unlock(pp); } continue; } /* * Is this page involved in some I/O? shared? * * The page_struct_lock need not be acquired to * examine these fields since the page has an * "exclusive" lock. */ if (pp->p_lckcnt != 0 || pp->p_cowcnt != 0) { page_unlock(pp); continue; } if (vp->v_type == VCHR) { panic("vp->v_type == VCHR"); /*NOTREACHED*/ } if (!page_try_demote_pages(pp)) { page_unlock(pp); continue; } /* * Check the modified bit. Leave the bits alone in hardware * (they will be modified if we do the putpage). */ mod = (hat_pagesync(pp, HAT_SYNC_DONTZERO | HAT_SYNC_STOPON_MOD) & P_MOD); if (mod) { offset = pp->p_offset; /* * Hold the vnode before releasing the page lock * to prevent it from being freed and re-used by * some other thread. */ VN_HOLD(vp); page_unlock(pp); /* * No error return is checked here. Callers such as * cpr deals with the dirty pages at the dump time * if this putpage fails. */ (void) VOP_PUTPAGE(vp, offset, PAGESIZE, B_INVAL, kcred, NULL); VN_RELE(vp); } else { /*LINTED: constant in conditional context*/ VN_DISPOSE(pp, B_INVAL, 0, kcred); } } while ((pp = page_next(pp)) != page0); if (nbusypages && retry++ < MAXRETRIES) { delay(1); goto top; } } /* * Replace the page "old" with the page "new" on the page hash and vnode lists * * the replacement must be done in place, ie the equivalent sequence: * * vp = old->p_vnode; * off = old->p_offset; * page_do_hashout(old) * page_do_hashin(new, vp, off) * * doesn't work, since * 1) if old is the only page on the vnode, the v_pages list has a window * where it looks empty. This will break file system assumptions. * and * 2) pvn_vplist_dirty() can't deal with pages moving on the v_pages list. */ static void page_do_relocate_hash(page_t *new, page_t *old) { page_t **hash_list; vnode_t *vp = old->p_vnode; kmutex_t *sep; ASSERT(PAGE_EXCL(old)); ASSERT(PAGE_EXCL(new)); ASSERT(vp != NULL); ASSERT(MUTEX_HELD(page_vnode_mutex(vp))); ASSERT(MUTEX_HELD(PAGE_HASH_MUTEX(PAGE_HASH_FUNC(vp, old->p_offset)))); /* * First find old page on the page hash list */ hash_list = &page_hash[PAGE_HASH_FUNC(vp, old->p_offset)]; for (;;) { if (*hash_list == old) break; if (*hash_list == NULL) { panic("page_do_hashout"); /*NOTREACHED*/ } hash_list = &(*hash_list)->p_hash; } /* * update new and replace old with new on the page hash list */ new->p_vnode = old->p_vnode; new->p_offset = old->p_offset; new->p_hash = old->p_hash; *hash_list = new; if ((new->p_vnode->v_flag & VISSWAP) != 0) PP_SETSWAP(new); /* * replace old with new on the vnode's page list */ if (old->p_vpnext == old) { new->p_vpnext = new; new->p_vpprev = new; } else { new->p_vpnext = old->p_vpnext; new->p_vpprev = old->p_vpprev; new->p_vpnext->p_vpprev = new; new->p_vpprev->p_vpnext = new; } if (vp->v_pages == old) vp->v_pages = new; /* * clear out the old page */ old->p_hash = NULL; old->p_vpnext = NULL; old->p_vpprev = NULL; old->p_vnode = NULL; PP_CLRSWAP(old); old->p_offset = (u_offset_t)-1; page_clr_all_props(old); /* * Wake up processes waiting for this page. The page's * identity has been changed, and is probably not the * desired page any longer. */ sep = page_se_mutex(old); mutex_enter(sep); old->p_selock &= ~SE_EWANTED; if (CV_HAS_WAITERS(&old->p_cv)) cv_broadcast(&old->p_cv); mutex_exit(sep); } /* * This function moves the identity of page "pp_old" to page "pp_new". * Both pages must be locked on entry. "pp_new" is free, has no identity, * and need not be hashed out from anywhere. */ void page_relocate_hash(page_t *pp_new, page_t *pp_old) { vnode_t *vp = pp_old->p_vnode; u_offset_t off = pp_old->p_offset; kmutex_t *phm, *vphm; /* * Rehash two pages */ ASSERT(PAGE_EXCL(pp_old)); ASSERT(PAGE_EXCL(pp_new)); ASSERT(vp != NULL); ASSERT(pp_new->p_vnode == NULL); /* * hashout then hashin while holding the mutexes */ phm = PAGE_HASH_MUTEX(PAGE_HASH_FUNC(vp, off)); mutex_enter(phm); vphm = page_vnode_mutex(vp); mutex_enter(vphm); page_do_relocate_hash(pp_new, pp_old); /* The following comment preserved from page_flip(). */ pp_new->p_fsdata = pp_old->p_fsdata; pp_old->p_fsdata = 0; mutex_exit(vphm); mutex_exit(phm); /* * The page_struct_lock need not be acquired for lckcnt and * cowcnt since the page has an "exclusive" lock. */ ASSERT(pp_new->p_lckcnt == 0); ASSERT(pp_new->p_cowcnt == 0); pp_new->p_lckcnt = pp_old->p_lckcnt; pp_new->p_cowcnt = pp_old->p_cowcnt; pp_old->p_lckcnt = pp_old->p_cowcnt = 0; } /* * Helper routine used to lock all remaining members of a * large page. The caller is responsible for passing in a locked * pp. If pp is a large page, then it succeeds in locking all the * remaining constituent pages or it returns with only the * original page locked. * * Returns 1 on success, 0 on failure. * * If success is returned this routine guarantees p_szc for all constituent * pages of a large page pp belongs to can't change. To achieve this we * recheck szc of pp after locking all constituent pages and retry if szc * changed (it could only decrease). Since hat_page_demote() needs an EXCL * lock on one of constituent pages it can't be running after all constituent * pages are locked. hat_page_demote() with a lock on a constituent page * outside of this large page (i.e. pp belonged to a larger large page) is * already done with all constituent pages of pp since the root's p_szc is * changed last. Therefore no need to synchronize with hat_page_demote() that * locked a constituent page outside of pp's current large page. */ #ifdef DEBUG uint32_t gpg_trylock_mtbf = 0; #endif int group_page_trylock(page_t *pp, se_t se) { page_t *tpp; pgcnt_t npgs, i, j; uint_t pszc = pp->p_szc; #ifdef DEBUG if (gpg_trylock_mtbf && !(gethrtime() % gpg_trylock_mtbf)) { return (0); } #endif if (pp != PP_GROUPLEADER(pp, pszc)) { return (0); } retry: ASSERT(PAGE_LOCKED_SE(pp, se)); ASSERT(!PP_ISFREE(pp)); if (pszc == 0) { return (1); } npgs = page_get_pagecnt(pszc); tpp = pp + 1; for (i = 1; i < npgs; i++, tpp++) { if (!page_trylock(tpp, se)) { tpp = pp + 1; for (j = 1; j < i; j++, tpp++) { page_unlock(tpp); } return (0); } } if (pp->p_szc != pszc) { ASSERT(pp->p_szc < pszc); ASSERT(pp->p_vnode != NULL && !PP_ISKAS(pp) && !IS_SWAPFSVP(pp->p_vnode)); tpp = pp + 1; for (i = 1; i < npgs; i++, tpp++) { page_unlock(tpp); } pszc = pp->p_szc; goto retry; } return (1); } void group_page_unlock(page_t *pp) { page_t *tpp; pgcnt_t npgs, i; ASSERT(PAGE_LOCKED(pp)); ASSERT(!PP_ISFREE(pp)); ASSERT(pp == PP_PAGEROOT(pp)); npgs = page_get_pagecnt(pp->p_szc); for (i = 1, tpp = pp + 1; i < npgs; i++, tpp++) { page_unlock(tpp); } } /* * returns * 0 : on success and *nrelocp is number of relocated PAGESIZE pages * ERANGE : this is not a base page * EBUSY : failure to get locks on the page/pages * ENOMEM : failure to obtain replacement pages * EAGAIN : OBP has not yet completed its boot-time handoff to the kernel * EIO : An error occurred while trying to copy the page data * * Return with all constituent members of target and replacement * SE_EXCL locked. It is the callers responsibility to drop the * locks. */ int do_page_relocate( page_t **target, page_t **replacement, int grouplock, spgcnt_t *nrelocp, lgrp_t *lgrp) { page_t *first_repl; page_t *repl; page_t *targ; page_t *pl = NULL; uint_t ppattr; pfn_t pfn, repl_pfn; uint_t szc; spgcnt_t npgs, i; int repl_contig = 0; uint_t flags = 0; spgcnt_t dofree = 0; *nrelocp = 0; #if defined(__sparc) /* * We need to wait till OBP has completed * its boot-time handoff of its resources to the kernel * before we allow page relocation */ if (page_relocate_ready == 0) { return (EAGAIN); } #endif /* * If this is not a base page, * just return with 0x0 pages relocated. */ targ = *target; ASSERT(PAGE_EXCL(targ)); ASSERT(!PP_ISFREE(targ)); szc = targ->p_szc; ASSERT(szc < mmu_page_sizes); VM_STAT_ADD(vmm_vmstats.ppr_reloc[szc]); pfn = targ->p_pagenum; if (pfn != PFN_BASE(pfn, szc)) { VM_STAT_ADD(vmm_vmstats.ppr_relocnoroot[szc]); return (ERANGE); } if ((repl = *replacement) != NULL && repl->p_szc >= szc) { repl_pfn = repl->p_pagenum; if (repl_pfn != PFN_BASE(repl_pfn, szc)) { VM_STAT_ADD(vmm_vmstats.ppr_reloc_replnoroot[szc]); return (ERANGE); } repl_contig = 1; } /* * We must lock all members of this large page or we cannot * relocate any part of it. */ if (grouplock != 0 && !group_page_trylock(targ, SE_EXCL)) { VM_STAT_ADD(vmm_vmstats.ppr_relocnolock[targ->p_szc]); return (EBUSY); } /* * reread szc it could have been decreased before * group_page_trylock() was done. */ szc = targ->p_szc; ASSERT(szc < mmu_page_sizes); VM_STAT_ADD(vmm_vmstats.ppr_reloc[szc]); ASSERT(pfn == PFN_BASE(pfn, szc)); npgs = page_get_pagecnt(targ->p_szc); if (repl == NULL) { dofree = npgs; /* Size of target page in MMU pages */ if (!page_create_wait(dofree, 0)) { if (grouplock != 0) { group_page_unlock(targ); } VM_STAT_ADD(vmm_vmstats.ppr_relocnomem[szc]); return (ENOMEM); } /* * seg kmem pages require that the target and replacement * page be the same pagesize. */ flags = (VN_ISKAS(targ->p_vnode)) ? PGR_SAMESZC : 0; repl = page_get_replacement_page(targ, lgrp, flags); if (repl == NULL) { if (grouplock != 0) { group_page_unlock(targ); } page_create_putback(dofree); VM_STAT_ADD(vmm_vmstats.ppr_relocnomem[szc]); return (ENOMEM); } } #ifdef DEBUG else { ASSERT(PAGE_LOCKED(repl)); } #endif /* DEBUG */ #if defined(__sparc) /* * Let hat_page_relocate() complete the relocation if it's kernel page */ if (VN_ISKAS(targ->p_vnode)) { *replacement = repl; if (hat_page_relocate(target, replacement, nrelocp) != 0) { if (grouplock != 0) { group_page_unlock(targ); } if (dofree) { *replacement = NULL; page_free_replacement_page(repl); page_create_putback(dofree); } VM_STAT_ADD(vmm_vmstats.ppr_krelocfail[szc]); return (EAGAIN); } VM_STAT_ADD(vmm_vmstats.ppr_relocok[szc]); return (0); } #else #if defined(lint) dofree = dofree; #endif #endif first_repl = repl; for (i = 0; i < npgs; i++) { ASSERT(PAGE_EXCL(targ)); ASSERT(targ->p_slckcnt == 0); ASSERT(repl->p_slckcnt == 0); (void) hat_pageunload(targ, HAT_FORCE_PGUNLOAD); ASSERT(hat_page_getshare(targ) == 0); ASSERT(!PP_ISFREE(targ)); ASSERT(targ->p_pagenum == (pfn + i)); ASSERT(repl_contig == 0 || repl->p_pagenum == (repl_pfn + i)); /* * Copy the page contents and attributes then * relocate the page in the page hash. */ if (ppcopy(targ, repl) == 0) { targ = *target; repl = first_repl; VM_STAT_ADD(vmm_vmstats.ppr_copyfail); if (grouplock != 0) { group_page_unlock(targ); } if (dofree) { *replacement = NULL; page_free_replacement_page(repl); page_create_putback(dofree); } return (EIO); } targ++; if (repl_contig != 0) { repl++; } else { repl = repl->p_next; } } repl = first_repl; targ = *target; for (i = 0; i < npgs; i++) { ppattr = hat_page_getattr(targ, (P_MOD | P_REF | P_RO)); page_clr_all_props(repl); page_set_props(repl, ppattr); page_relocate_hash(repl, targ); ASSERT(hat_page_getshare(targ) == 0); ASSERT(hat_page_getshare(repl) == 0); /* * Now clear the props on targ, after the * page_relocate_hash(), they no longer * have any meaning. */ page_clr_all_props(targ); ASSERT(targ->p_next == targ); ASSERT(targ->p_prev == targ); page_list_concat(&pl, &targ); targ++; if (repl_contig != 0) { repl++; } else { repl = repl->p_next; } } /* assert that we have come full circle with repl */ ASSERT(repl_contig == 1 || first_repl == repl); *target = pl; if (*replacement == NULL) { ASSERT(first_repl == repl); *replacement = repl; } VM_STAT_ADD(vmm_vmstats.ppr_relocok[szc]); *nrelocp = npgs; return (0); } /* * On success returns 0 and *nrelocp the number of PAGESIZE pages relocated. */ int page_relocate( page_t **target, page_t **replacement, int grouplock, int freetarget, spgcnt_t *nrelocp, lgrp_t *lgrp) { spgcnt_t ret; /* do_page_relocate returns 0 on success or errno value */ ret = do_page_relocate(target, replacement, grouplock, nrelocp, lgrp); if (ret != 0 || freetarget == 0) { return (ret); } if (*nrelocp == 1) { ASSERT(*target != NULL); page_free(*target, 1); } else { page_t *tpp = *target; uint_t szc = tpp->p_szc; pgcnt_t npgs = page_get_pagecnt(szc); ASSERT(npgs > 1); ASSERT(szc != 0); do { ASSERT(PAGE_EXCL(tpp)); ASSERT(!hat_page_is_mapped(tpp)); ASSERT(tpp->p_szc == szc); PP_SETFREE(tpp); PP_SETAGED(tpp); npgs--; } while ((tpp = tpp->p_next) != *target); ASSERT(npgs == 0); page_list_add_pages(*target, 0); npgs = page_get_pagecnt(szc); page_create_putback(npgs); } return (ret); } /* * it is up to the caller to deal with pcf accounting. */ void page_free_replacement_page(page_t *pplist) { page_t *pp; while (pplist != NULL) { /* * pp_targ is a linked list. */ pp = pplist; if (pp->p_szc == 0) { page_sub(&pplist, pp); page_clr_all_props(pp); PP_SETFREE(pp); PP_SETAGED(pp); page_list_add(pp, PG_FREE_LIST | PG_LIST_TAIL); page_unlock(pp); VM_STAT_ADD(pagecnt.pc_free_replacement_page[0]); } else { spgcnt_t curnpgs = page_get_pagecnt(pp->p_szc); page_t *tpp; page_list_break(&pp, &pplist, curnpgs); tpp = pp; do { ASSERT(PAGE_EXCL(tpp)); ASSERT(!hat_page_is_mapped(tpp)); page_clr_all_props(tpp); PP_SETFREE(tpp); PP_SETAGED(tpp); } while ((tpp = tpp->p_next) != pp); page_list_add_pages(pp, 0); VM_STAT_ADD(pagecnt.pc_free_replacement_page[1]); } } } /* * Relocate target to non-relocatable replacement page. */ int page_relocate_cage(page_t **target, page_t **replacement) { page_t *tpp, *rpp; spgcnt_t pgcnt, npgs; int result; tpp = *target; ASSERT(PAGE_EXCL(tpp)); ASSERT(tpp->p_szc == 0); pgcnt = btop(page_get_pagesize(tpp->p_szc)); do { (void) page_create_wait(pgcnt, PG_WAIT | PG_NORELOC); rpp = page_get_replacement_page(tpp, NULL, PGR_NORELOC); if (rpp == NULL) { page_create_putback(pgcnt); kcage_cageout_wakeup(); } } while (rpp == NULL); ASSERT(PP_ISNORELOC(rpp)); result = page_relocate(&tpp, &rpp, 0, 1, &npgs, NULL); if (result == 0) { *replacement = rpp; if (pgcnt != npgs) panic("page_relocate_cage: partial relocation"); } return (result); } /* * Release the page lock on a page, place on cachelist * tail if no longer mapped. Caller can let us know if * the page is known to be clean. */ int page_release(page_t *pp, int checkmod) { int status; ASSERT(PAGE_LOCKED(pp) && !PP_ISFREE(pp) && (pp->p_vnode != NULL)); if (!hat_page_is_mapped(pp) && !IS_SWAPVP(pp->p_vnode) && ((PAGE_SHARED(pp) && page_tryupgrade(pp)) || PAGE_EXCL(pp)) && pp->p_lckcnt == 0 && pp->p_cowcnt == 0 && !hat_page_is_mapped(pp)) { /* * If page is modified, unlock it * * (p_nrm & P_MOD) bit has the latest stuff because: * (1) We found that this page doesn't have any mappings * _after_ holding SE_EXCL and * (2) We didn't drop SE_EXCL lock after the check in (1) */ if (checkmod && hat_ismod(pp)) { page_unlock(pp); status = PGREL_MOD; } else { /*LINTED: constant in conditional context*/ VN_DISPOSE(pp, B_FREE, 0, kcred); status = PGREL_CLEAN; } } else { page_unlock(pp); status = PGREL_NOTREL; } return (status); } /* * Given a constituent page, try to demote the large page on the freelist. * * Returns nonzero if the page could be demoted successfully. Returns with * the constituent page still locked. */ int page_try_demote_free_pages(page_t *pp) { page_t *rootpp = pp; pfn_t pfn = page_pptonum(pp); spgcnt_t npgs; uint_t szc = pp->p_szc; ASSERT(PP_ISFREE(pp)); ASSERT(PAGE_EXCL(pp)); /* * Adjust rootpp and lock it, if `pp' is not the base * constituent page. */ npgs = page_get_pagecnt(pp->p_szc); if (npgs == 1) { return (0); } if (!IS_P2ALIGNED(pfn, npgs)) { pfn = P2ALIGN(pfn, npgs); rootpp = page_numtopp_nolock(pfn); } if (pp != rootpp && !page_trylock(rootpp, SE_EXCL)) { return (0); } if (rootpp->p_szc != szc) { if (pp != rootpp) page_unlock(rootpp); return (0); } page_demote_free_pages(rootpp); if (pp != rootpp) page_unlock(rootpp); ASSERT(PP_ISFREE(pp)); ASSERT(PAGE_EXCL(pp)); return (1); } /* * Given a constituent page, try to demote the large page. * * Returns nonzero if the page could be demoted successfully. Returns with * the constituent page still locked. */ int page_try_demote_pages(page_t *pp) { page_t *tpp, *rootpp = pp; pfn_t pfn = page_pptonum(pp); spgcnt_t i, npgs; uint_t szc = pp->p_szc; vnode_t *vp = pp->p_vnode; ASSERT(PAGE_EXCL(pp)); VM_STAT_ADD(pagecnt.pc_try_demote_pages[0]); if (pp->p_szc == 0) { VM_STAT_ADD(pagecnt.pc_try_demote_pages[1]); return (1); } if (vp != NULL && !IS_SWAPFSVP(vp) && !VN_ISKAS(vp)) { VM_STAT_ADD(pagecnt.pc_try_demote_pages[2]); page_demote_vp_pages(pp); ASSERT(pp->p_szc == 0); return (1); } /* * Adjust rootpp if passed in is not the base * constituent page. */ npgs = page_get_pagecnt(pp->p_szc); ASSERT(npgs > 1); if (!IS_P2ALIGNED(pfn, npgs)) { pfn = P2ALIGN(pfn, npgs); rootpp = page_numtopp_nolock(pfn); VM_STAT_ADD(pagecnt.pc_try_demote_pages[3]); ASSERT(rootpp->p_vnode != NULL); ASSERT(rootpp->p_szc == szc); } /* * We can't demote kernel pages since we can't hat_unload() * the mappings. */ if (VN_ISKAS(rootpp->p_vnode)) return (0); /* * Attempt to lock all constituent pages except the page passed * in since it's already locked. */ for (tpp = rootpp, i = 0; i < npgs; i++, tpp++) { ASSERT(!PP_ISFREE(tpp)); ASSERT(tpp->p_vnode != NULL); if (tpp != pp && !page_trylock(tpp, SE_EXCL)) break; ASSERT(tpp->p_szc == rootpp->p_szc); ASSERT(page_pptonum(tpp) == page_pptonum(rootpp) + i); } /* * If we failed to lock them all then unlock what we have * locked so far and bail. */ if (i < npgs) { tpp = rootpp; while (i-- > 0) { if (tpp != pp) page_unlock(tpp); tpp++; } VM_STAT_ADD(pagecnt.pc_try_demote_pages[4]); return (0); } for (tpp = rootpp, i = 0; i < npgs; i++, tpp++) { ASSERT(PAGE_EXCL(tpp)); ASSERT(tpp->p_slckcnt == 0); (void) hat_pageunload(tpp, HAT_FORCE_PGUNLOAD); tpp->p_szc = 0; } /* * Unlock all pages except the page passed in. */ for (tpp = rootpp, i = 0; i < npgs; i++, tpp++) { ASSERT(!hat_page_is_mapped(tpp)); if (tpp != pp) page_unlock(tpp); } VM_STAT_ADD(pagecnt.pc_try_demote_pages[5]); return (1); } /* * Called by page_free() and page_destroy() to demote the page size code * (p_szc) to 0 (since we can't just put a single PAGESIZE page with non zero * p_szc on free list, neither can we just clear p_szc of a single page_t * within a large page since it will break other code that relies on p_szc * being the same for all page_t's of a large page). Anonymous pages should * never end up here because anon_map_getpages() cannot deal with p_szc * changes after a single constituent page is locked. While anonymous or * kernel large pages are demoted or freed the entire large page at a time * with all constituent pages locked EXCL for the file system pages we * have to be able to demote a large page (i.e. decrease all constituent pages * p_szc) with only just an EXCL lock on one of constituent pages. The reason * we can easily deal with anonymous page demotion the entire large page at a * time is that those operation originate at address space level and concern * the entire large page region with actual demotion only done when pages are * not shared with any other processes (therefore we can always get EXCL lock * on all anonymous constituent pages after clearing segment page * cache). However file system pages can be truncated or invalidated at a * PAGESIZE level from the file system side and end up in page_free() or * page_destroy() (we also allow only part of the large page to be SOFTLOCKed * and therefore pageout should be able to demote a large page by EXCL locking * any constituent page that is not under SOFTLOCK). In those cases we cannot * rely on being able to lock EXCL all constituent pages. * * To prevent szc changes on file system pages one has to lock all constituent * pages at least SHARED (or call page_szc_lock()). The only subsystem that * doesn't rely on locking all constituent pages (or using page_szc_lock()) to * prevent szc changes is hat layer that uses its own page level mlist * locks. hat assumes that szc doesn't change after mlist lock for a page is * taken. Therefore we need to change szc under hat level locks if we only * have an EXCL lock on a single constituent page and hat still references any * of constituent pages. (Note we can't "ignore" hat layer by simply * hat_pageunload() all constituent pages without having EXCL locks on all of * constituent pages). We use hat_page_demote() call to safely demote szc of * all constituent pages under hat locks when we only have an EXCL lock on one * of constituent pages. * * This routine calls page_szc_lock() before calling hat_page_demote() to * allow segvn in one special case not to lock all constituent pages SHARED * before calling hat_memload_array() that relies on p_szc not changing even * before hat level mlist lock is taken. In that case segvn uses * page_szc_lock() to prevent hat_page_demote() changing p_szc values. * * Anonymous or kernel page demotion still has to lock all pages exclusively * and do hat_pageunload() on all constituent pages before demoting the page * therefore there's no need for anonymous or kernel page demotion to use * hat_page_demote() mechanism. * * hat_page_demote() removes all large mappings that map pp and then decreases * p_szc starting from the last constituent page of the large page. By working * from the tail of a large page in pfn decreasing order allows one looking at * the root page to know that hat_page_demote() is done for root's szc area. * e.g. if a root page has szc 1 one knows it only has to lock all constituent * pages within szc 1 area to prevent szc changes because hat_page_demote() * that started on this page when it had szc > 1 is done for this szc 1 area. * * We are guaranteed that all constituent pages of pp's large page belong to * the same vnode with the consecutive offsets increasing in the direction of * the pfn i.e. the identity of constituent pages can't change until their * p_szc is decreased. Therefore it's safe for hat_page_demote() to remove * large mappings to pp even though we don't lock any constituent page except * pp (i.e. we won't unload e.g. kernel locked page). */ static void page_demote_vp_pages(page_t *pp) { kmutex_t *mtx; ASSERT(PAGE_EXCL(pp)); ASSERT(!PP_ISFREE(pp)); ASSERT(pp->p_vnode != NULL); ASSERT(!IS_SWAPFSVP(pp->p_vnode)); ASSERT(!PP_ISKAS(pp)); VM_STAT_ADD(pagecnt.pc_demote_pages[0]); mtx = page_szc_lock(pp); if (mtx != NULL) { hat_page_demote(pp); mutex_exit(mtx); } ASSERT(pp->p_szc == 0); } /* * Mark any existing pages for migration in the given range */ void page_mark_migrate(struct seg *seg, caddr_t addr, size_t len, struct anon_map *amp, ulong_t anon_index, vnode_t *vp, u_offset_t vnoff, int rflag) { struct anon *ap; vnode_t *curvp; lgrp_t *from; pgcnt_t nlocked; u_offset_t off; pfn_t pfn; size_t pgsz; size_t segpgsz; pgcnt_t pages; uint_t pszc; page_t *pp0, *pp; caddr_t va; ulong_t an_idx; anon_sync_obj_t cookie; ASSERT(seg->s_as && AS_LOCK_HELD(seg->s_as)); /* * Don't do anything if don't need to do lgroup optimizations * on this system */ if (!lgrp_optimizations()) return; /* * Align address and length to (potentially large) page boundary */ segpgsz = page_get_pagesize(seg->s_szc); addr = (caddr_t)P2ALIGN((uintptr_t)addr, segpgsz); if (rflag) len = P2ROUNDUP(len, segpgsz); /* * Do one (large) page at a time */ va = addr; while (va < addr + len) { /* * Lookup (root) page for vnode and offset corresponding to * this virtual address * Try anonmap first since there may be copy-on-write * pages, but initialize vnode pointer and offset using * vnode arguments just in case there isn't an amp. */ curvp = vp; off = vnoff + va - seg->s_base; if (amp) { ANON_LOCK_ENTER(&->a_rwlock, RW_READER); an_idx = anon_index + seg_page(seg, va); anon_array_enter(amp, an_idx, &cookie); ap = anon_get_ptr(amp->ahp, an_idx); if (ap) swap_xlate(ap, &curvp, &off); anon_array_exit(&cookie); ANON_LOCK_EXIT(&->a_rwlock); } pp = NULL; if (curvp) pp = page_lookup(curvp, off, SE_SHARED); /* * If there isn't a page at this virtual address, * skip to next page */ if (pp == NULL) { va += PAGESIZE; continue; } /* * Figure out which lgroup this page is in for kstats */ pfn = page_pptonum(pp); from = lgrp_pfn_to_lgrp(pfn); /* * Get page size, and round up and skip to next page boundary * if unaligned address */ pszc = pp->p_szc; pgsz = page_get_pagesize(pszc); pages = btop(pgsz); if (!IS_P2ALIGNED(va, pgsz) || !IS_P2ALIGNED(pfn, pages) || pgsz > segpgsz) { pgsz = MIN(pgsz, segpgsz); page_unlock(pp); pages = btop(P2END((uintptr_t)va, pgsz) - (uintptr_t)va); va = (caddr_t)P2END((uintptr_t)va, pgsz); lgrp_stat_add(from->lgrp_id, LGRP_PMM_FAIL_PGS, pages); continue; } /* * Upgrade to exclusive lock on page */ if (!page_tryupgrade(pp)) { page_unlock(pp); va += pgsz; lgrp_stat_add(from->lgrp_id, LGRP_PMM_FAIL_PGS, btop(pgsz)); continue; } pp0 = pp++; nlocked = 1; /* * Lock constituent pages if this is large page */ if (pages > 1) { /* * Lock all constituents except root page, since it * should be locked already. */ for (; nlocked < pages; nlocked++) { if (!page_trylock(pp, SE_EXCL)) { break; } if (PP_ISFREE(pp) || pp->p_szc != pszc) { /* * hat_page_demote() raced in with us. */ ASSERT(!IS_SWAPFSVP(curvp)); page_unlock(pp); break; } pp++; } } /* * If all constituent pages couldn't be locked, * unlock pages locked so far and skip to next page. */ if (nlocked < pages) { while (pp0 < pp) { page_unlock(pp0++); } va += pgsz; lgrp_stat_add(from->lgrp_id, LGRP_PMM_FAIL_PGS, btop(pgsz)); continue; } /* * hat_page_demote() can no longer happen * since last cons page had the right p_szc after * all cons pages were locked. all cons pages * should now have the same p_szc. */ /* * All constituent pages locked successfully, so mark * large page for migration and unload the mappings of * constituent pages, so a fault will occur on any part of the * large page */ PP_SETMIGRATE(pp0); while (pp0 < pp) { (void) hat_pageunload(pp0, HAT_FORCE_PGUNLOAD); ASSERT(hat_page_getshare(pp0) == 0); page_unlock(pp0++); } lgrp_stat_add(from->lgrp_id, LGRP_PMM_PGS, nlocked); va += pgsz; } } /* * Migrate any pages that have been marked for migration in the given range */ void page_migrate( struct seg *seg, caddr_t addr, page_t **ppa, pgcnt_t npages) { lgrp_t *from; lgrp_t *to; page_t *newpp; page_t *pp; pfn_t pfn; size_t pgsz; spgcnt_t page_cnt; spgcnt_t i; uint_t pszc; ASSERT(seg->s_as && AS_LOCK_HELD(seg->s_as)); while (npages > 0) { pp = *ppa; pszc = pp->p_szc; pgsz = page_get_pagesize(pszc); page_cnt = btop(pgsz); /* * Check to see whether this page is marked for migration * * Assume that root page of large page is marked for * migration and none of the other constituent pages * are marked. This really simplifies clearing the * migrate bit by not having to clear it from each * constituent page. * * note we don't want to relocate an entire large page if * someone is only using one subpage. */ if (npages < page_cnt) break; /* * Is it marked for migration? */ if (!PP_ISMIGRATE(pp)) goto next; /* * Determine lgroups that page is being migrated between */ pfn = page_pptonum(pp); if (!IS_P2ALIGNED(pfn, page_cnt)) { break; } from = lgrp_pfn_to_lgrp(pfn); to = lgrp_mem_choose(seg, addr, pgsz); /* * Need to get exclusive lock's to migrate */ for (i = 0; i < page_cnt; i++) { ASSERT(PAGE_LOCKED(ppa[i])); if (page_pptonum(ppa[i]) != pfn + i || ppa[i]->p_szc != pszc) { break; } if (!page_tryupgrade(ppa[i])) { lgrp_stat_add(from->lgrp_id, LGRP_PM_FAIL_LOCK_PGS, page_cnt); break; } /* * Check to see whether we are trying to migrate * page to lgroup where it is allocated already. * If so, clear the migrate bit and skip to next * page. */ if (i == 0 && to == from) { PP_CLRMIGRATE(ppa[0]); page_downgrade(ppa[0]); goto next; } } /* * If all constituent pages couldn't be locked, * unlock pages locked so far and skip to next page. */ if (i != page_cnt) { while (--i != -1) { page_downgrade(ppa[i]); } goto next; } (void) page_create_wait(page_cnt, PG_WAIT); newpp = page_get_replacement_page(pp, to, PGR_SAMESZC); if (newpp == NULL) { page_create_putback(page_cnt); for (i = 0; i < page_cnt; i++) { page_downgrade(ppa[i]); } lgrp_stat_add(to->lgrp_id, LGRP_PM_FAIL_ALLOC_PGS, page_cnt); goto next; } ASSERT(newpp->p_szc == pszc); /* * Clear migrate bit and relocate page */ PP_CLRMIGRATE(pp); if (page_relocate(&pp, &newpp, 0, 1, &page_cnt, to)) { panic("page_migrate: page_relocate failed"); } ASSERT(page_cnt * PAGESIZE == pgsz); /* * Keep stats for number of pages migrated from and to * each lgroup */ lgrp_stat_add(from->lgrp_id, LGRP_PM_SRC_PGS, page_cnt); lgrp_stat_add(to->lgrp_id, LGRP_PM_DEST_PGS, page_cnt); /* * update the page_t array we were passed in and * unlink constituent pages of a large page. */ for (i = 0; i < page_cnt; ++i, ++pp) { ASSERT(PAGE_EXCL(newpp)); ASSERT(newpp->p_szc == pszc); ppa[i] = newpp; pp = newpp; page_sub(&newpp, pp); page_downgrade(pp); } ASSERT(newpp == NULL); next: addr += pgsz; ppa += page_cnt; npages -= page_cnt; } } uint_t page_reclaim_maxcnt = 60; /* max total iterations */ uint_t page_reclaim_nofree_maxcnt = 3; /* max iterations without progress */ /* * Reclaim/reserve availrmem for npages. * If there is not enough memory start reaping seg, kmem caches. * Start pageout scanner (via page_needfree()). * Exit after ~ MAX_CNT s regardless of how much memory has been released. * Note: There is no guarantee that any availrmem will be freed as * this memory typically is locked (kernel heap) or reserved for swap. * Also due to memory fragmentation kmem allocator may not be able * to free any memory (single user allocated buffer will prevent * freeing slab or a page). */ int page_reclaim_mem(pgcnt_t npages, pgcnt_t epages, int adjust) { int i = 0; int i_nofree = 0; int ret = 0; pgcnt_t deficit; pgcnt_t old_availrmem = 0; mutex_enter(&freemem_lock); while (availrmem < tune.t_minarmem + npages + epages && i++ < page_reclaim_maxcnt) { /* ensure we made some progress in the last few iterations */ if (old_availrmem < availrmem) { old_availrmem = availrmem; i_nofree = 0; } else if (i_nofree++ >= page_reclaim_nofree_maxcnt) { break; } deficit = tune.t_minarmem + npages + epages - availrmem; mutex_exit(&freemem_lock); page_needfree(deficit); kmem_reap(); delay(hz); page_needfree(-(spgcnt_t)deficit); mutex_enter(&freemem_lock); } if (adjust && (availrmem >= tune.t_minarmem + npages + epages)) { availrmem -= npages; ret = 1; } mutex_exit(&freemem_lock); return (ret); } /* * Search the memory segments to locate the desired page. Within a * segment, pages increase linearly with one page structure per * physical page frame (size PAGESIZE). The search begins * with the segment that was accessed last, to take advantage of locality. * If the hint misses, we start from the beginning of the sorted memseg list */ /* * Some data structures for pfn to pp lookup. */ ulong_t mhash_per_slot; struct memseg *memseg_hash[N_MEM_SLOTS]; page_t * page_numtopp_nolock(pfn_t pfnum) { struct memseg *seg; page_t *pp; vm_cpu_data_t *vc; /* * We need to disable kernel preemption while referencing the * cpu_vm_data field in order to prevent us from being switched to * another cpu and trying to reference it after it has been freed. * This will keep us on cpu and prevent it from being removed while * we are still on it. * * We may be caching a memseg in vc_pnum_memseg/vc_pnext_memseg * which is being resued by DR who will flush those references * before modifying the reused memseg. See memseg_cpu_vm_flush(). */ kpreempt_disable(); vc = CPU->cpu_vm_data; ASSERT(vc != NULL); MEMSEG_STAT_INCR(nsearch); /* Try last winner first */ if (((seg = vc->vc_pnum_memseg) != NULL) && (pfnum >= seg->pages_base) && (pfnum < seg->pages_end)) { MEMSEG_STAT_INCR(nlastwon); pp = seg->pages + (pfnum - seg->pages_base); if (pp->p_pagenum == pfnum) { kpreempt_enable(); return ((page_t *)pp); } } /* Else Try hash */ if (((seg = memseg_hash[MEMSEG_PFN_HASH(pfnum)]) != NULL) && (pfnum >= seg->pages_base) && (pfnum < seg->pages_end)) { MEMSEG_STAT_INCR(nhashwon); vc->vc_pnum_memseg = seg; pp = seg->pages + (pfnum - seg->pages_base); if (pp->p_pagenum == pfnum) { kpreempt_enable(); return ((page_t *)pp); } } /* Else Brute force */ for (seg = memsegs; seg != NULL; seg = seg->next) { if (pfnum >= seg->pages_base && pfnum < seg->pages_end) { vc->vc_pnum_memseg = seg; pp = seg->pages + (pfnum - seg->pages_base); if (pp->p_pagenum == pfnum) { kpreempt_enable(); return ((page_t *)pp); } } } vc->vc_pnum_memseg = NULL; kpreempt_enable(); MEMSEG_STAT_INCR(nnotfound); return ((page_t *)NULL); } struct memseg * page_numtomemseg_nolock(pfn_t pfnum) { struct memseg *seg; page_t *pp; /* * We may be caching a memseg in vc_pnum_memseg/vc_pnext_memseg * which is being resued by DR who will flush those references * before modifying the reused memseg. See memseg_cpu_vm_flush(). */ kpreempt_disable(); /* Try hash */ if (((seg = memseg_hash[MEMSEG_PFN_HASH(pfnum)]) != NULL) && (pfnum >= seg->pages_base) && (pfnum < seg->pages_end)) { pp = seg->pages + (pfnum - seg->pages_base); if (pp->p_pagenum == pfnum) { kpreempt_enable(); return (seg); } } /* Else Brute force */ for (seg = memsegs; seg != NULL; seg = seg->next) { if (pfnum >= seg->pages_base && pfnum < seg->pages_end) { pp = seg->pages + (pfnum - seg->pages_base); if (pp->p_pagenum == pfnum) { kpreempt_enable(); return (seg); } } } kpreempt_enable(); return ((struct memseg *)NULL); } /* * Given a page and a count return the page struct that is * n structs away from the current one in the global page * list. * * This function wraps to the first page upon * reaching the end of the memseg list. */ page_t * page_nextn(page_t *pp, ulong_t n) { struct memseg *seg; page_t *ppn; vm_cpu_data_t *vc; /* * We need to disable kernel preemption while referencing the * cpu_vm_data field in order to prevent us from being switched to * another cpu and trying to reference it after it has been freed. * This will keep us on cpu and prevent it from being removed while * we are still on it. * * We may be caching a memseg in vc_pnum_memseg/vc_pnext_memseg * which is being resued by DR who will flush those references * before modifying the reused memseg. See memseg_cpu_vm_flush(). */ kpreempt_disable(); vc = (vm_cpu_data_t *)CPU->cpu_vm_data; ASSERT(vc != NULL); if (((seg = vc->vc_pnext_memseg) == NULL) || (seg->pages_base == seg->pages_end) || !(pp >= seg->pages && pp < seg->epages)) { for (seg = memsegs; seg; seg = seg->next) { if (pp >= seg->pages && pp < seg->epages) break; } if (seg == NULL) { /* Memory delete got in, return something valid. */ /* TODO: fix me. */ seg = memsegs; pp = seg->pages; } } /* check for wraparound - possible if n is large */ while ((ppn = (pp + n)) >= seg->epages || ppn < pp) { n -= seg->epages - pp; seg = seg->next; if (seg == NULL) seg = memsegs; pp = seg->pages; } vc->vc_pnext_memseg = seg; kpreempt_enable(); return (ppn); } /* * Initialize for a loop using page_next_scan_large(). */ page_t * page_next_scan_init(void **cookie) { ASSERT(cookie != NULL); *cookie = (void *)memsegs; return ((page_t *)memsegs->pages); } /* * Return the next page in a scan of page_t's, assuming we want * to skip over sub-pages within larger page sizes. * * The cookie is used to keep track of the current memseg. */ page_t * page_next_scan_large( page_t *pp, ulong_t *n, void **cookie) { struct memseg *seg = (struct memseg *)*cookie; page_t *new_pp; ulong_t cnt; pfn_t pfn; /* * get the count of page_t's to skip based on the page size */ ASSERT(pp != NULL); if (pp->p_szc == 0) { cnt = 1; } else { pfn = page_pptonum(pp); cnt = page_get_pagecnt(pp->p_szc); cnt -= pfn & (cnt - 1); } *n += cnt; new_pp = pp + cnt; /* * Catch if we went past the end of the current memory segment. If so, * just move to the next segment with pages. */ if (new_pp >= seg->epages || seg->pages_base == seg->pages_end) { do { seg = seg->next; if (seg == NULL) seg = memsegs; } while (seg->pages_base == seg->pages_end); new_pp = seg->pages; *cookie = (void *)seg; } return (new_pp); } /* * Returns next page in list. Note: this function wraps * to the first page in the list upon reaching the end * of the list. Callers should be aware of this fact. */ /* We should change this be a #define */ page_t * page_next(page_t *pp) { return (page_nextn(pp, 1)); } page_t * page_first() { return ((page_t *)memsegs->pages); } /* * This routine is called at boot with the initial memory configuration * and when memory is added or removed. */ void build_pfn_hash() { pfn_t cur; pgcnt_t index; struct memseg *pseg; int i; /* * Clear memseg_hash array. * Since memory add/delete is designed to operate concurrently * with normal operation, the hash rebuild must be able to run * concurrently with page_numtopp_nolock(). To support this * functionality, assignments to memseg_hash array members must * be done atomically. * * NOTE: bzero() does not currently guarantee this for kernel * threads, and cannot be used here. */ for (i = 0; i < N_MEM_SLOTS; i++) memseg_hash[i] = NULL; hat_kpm_mseghash_clear(N_MEM_SLOTS); /* * Physmax is the last valid pfn. */ mhash_per_slot = (physmax + 1) >> MEM_HASH_SHIFT; for (pseg = memsegs; pseg != NULL; pseg = pseg->next) { index = MEMSEG_PFN_HASH(pseg->pages_base); cur = pseg->pages_base; do { if (index >= N_MEM_SLOTS) index = MEMSEG_PFN_HASH(cur); if (memseg_hash[index] == NULL || memseg_hash[index]->pages_base > pseg->pages_base) { memseg_hash[index] = pseg; hat_kpm_mseghash_update(index, pseg); } cur += mhash_per_slot; index++; } while (cur < pseg->pages_end); } } /* * Return the pagenum for the pp */ pfn_t page_pptonum(page_t *pp) { return (pp->p_pagenum); } /* * interface to the referenced and modified etc bits * in the PSM part of the page struct * when no locking is desired. */ void page_set_props(page_t *pp, uint_t flags) { ASSERT((flags & ~(P_MOD | P_REF | P_RO)) == 0); pp->p_nrm |= (uchar_t)flags; } void page_clr_all_props(page_t *pp) { pp->p_nrm = 0; } /* * Clear p_lckcnt and p_cowcnt, adjusting freemem if required. */ int page_clear_lck_cow(page_t *pp, int adjust) { int f_amount; ASSERT(PAGE_EXCL(pp)); /* * The page_struct_lock need not be acquired here since * we require the caller hold the page exclusively locked. */ f_amount = 0; if (pp->p_lckcnt) { f_amount = 1; pp->p_lckcnt = 0; } if (pp->p_cowcnt) { f_amount += pp->p_cowcnt; pp->p_cowcnt = 0; } if (adjust && f_amount) { mutex_enter(&freemem_lock); availrmem += f_amount; mutex_exit(&freemem_lock); } return (f_amount); } /* * The following functions is called from free_vp_pages() * for an inexact estimate of a newly free'd page... */ ulong_t page_share_cnt(page_t *pp) { return (hat_page_getshare(pp)); } int page_isshared(page_t *pp) { return (hat_page_checkshare(pp, 1)); } int page_isfree(page_t *pp) { return (PP_ISFREE(pp)); } int page_isref(page_t *pp) { return (hat_page_getattr(pp, P_REF)); } int page_ismod(page_t *pp) { return (hat_page_getattr(pp, P_MOD)); } /* * The following code all currently relates to the page capture logic: * * This logic is used for cases where there is a desire to claim a certain * physical page in the system for the caller. As it may not be possible * to capture the page immediately, the p_toxic bits are used in the page * structure to indicate that someone wants to capture this page. When the * page gets unlocked, the toxic flag will be noted and an attempt to capture * the page will be made. If it is successful, the original callers callback * will be called with the page to do with it what they please. * * There is also an async thread which wakes up to attempt to capture * pages occasionally which have the capture bit set. All of the pages which * need to be captured asynchronously have been inserted into the * page_capture_hash and thus this thread walks that hash list. Items in the * hash have an expiration time so this thread handles that as well by removing * the item from the hash if it has expired. * * Some important things to note are: * - if the PR_CAPTURE bit is set on a page, then the page is in the * page_capture_hash. The page_capture_hash_head.pchh_mutex is needed * to set and clear this bit, and while the lock is held is the only time * you can add or remove an entry from the hash. * - the PR_CAPTURE bit can only be set and cleared while holding the * page_capture_hash_head.pchh_mutex * - the t_flag field of the thread struct is used with the T_CAPTURING * flag to prevent recursion while dealing with large pages. * - pages which need to be retired never expire on the page_capture_hash. */ static void page_capture_thread(void); static kthread_t *pc_thread_id; kcondvar_t pc_cv; static kmutex_t pc_thread_mutex; static clock_t pc_thread_shortwait; static clock_t pc_thread_longwait; static int pc_thread_retry; struct page_capture_callback pc_cb[PC_NUM_CALLBACKS]; /* Note that this is a circular linked list */ typedef struct page_capture_hash_bucket { page_t *pp; uchar_t szc; uchar_t pri; uint_t flags; clock_t expires; /* lbolt at which this request expires. */ void *datap; /* Cached data passed in for callback */ struct page_capture_hash_bucket *next; struct page_capture_hash_bucket *prev; } page_capture_hash_bucket_t; #define PC_PRI_HI 0 /* capture now */ #define PC_PRI_LO 1 /* capture later */ #define PC_NUM_PRI 2 #define PAGE_CAPTURE_PRIO(pp) (PP_ISRAF(pp) ? PC_PRI_LO : PC_PRI_HI) /* * Each hash bucket will have it's own mutex and two lists which are: * active (0): represents requests which have not been processed by * the page_capture async thread yet. * walked (1): represents requests which have been processed by the * page_capture async thread within it's given walk of this bucket. * * These are all needed so that we can synchronize all async page_capture * events. When the async thread moves to a new bucket, it will append the * walked list to the active list and walk each item one at a time, moving it * from the active list to the walked list. Thus if there is an async request * outstanding for a given page, it will always be in one of the two lists. * New requests will always be added to the active list. * If we were not able to capture a page before the request expired, we'd free * up the request structure which would indicate to page_capture that there is * no longer a need for the given page, and clear the PR_CAPTURE flag if * possible. */ typedef struct page_capture_hash_head { kmutex_t pchh_mutex; uint_t num_pages[PC_NUM_PRI]; page_capture_hash_bucket_t lists[2]; /* sentinel nodes */ } page_capture_hash_head_t; #ifdef DEBUG #define NUM_PAGE_CAPTURE_BUCKETS 4 #else #define NUM_PAGE_CAPTURE_BUCKETS 64 #endif page_capture_hash_head_t page_capture_hash[NUM_PAGE_CAPTURE_BUCKETS]; /* for now use a very simple hash based upon the size of a page struct */ #define PAGE_CAPTURE_HASH(pp) \ ((int)(((uintptr_t)pp >> 7) & (NUM_PAGE_CAPTURE_BUCKETS - 1))) extern pgcnt_t swapfs_minfree; int page_trycapture(page_t *pp, uint_t szc, uint_t flags, void *datap); /* * a callback function is required for page capture requests. */ void page_capture_register_callback(uint_t index, clock_t duration, int (*cb_func)(page_t *, void *, uint_t)) { ASSERT(pc_cb[index].cb_active == 0); ASSERT(cb_func != NULL); rw_enter(&pc_cb[index].cb_rwlock, RW_WRITER); pc_cb[index].duration = duration; pc_cb[index].cb_func = cb_func; pc_cb[index].cb_active = 1; rw_exit(&pc_cb[index].cb_rwlock); } void page_capture_unregister_callback(uint_t index) { int i, j; struct page_capture_hash_bucket *bp1; struct page_capture_hash_bucket *bp2; struct page_capture_hash_bucket *head = NULL; uint_t flags = (1 << index); rw_enter(&pc_cb[index].cb_rwlock, RW_WRITER); ASSERT(pc_cb[index].cb_active == 1); pc_cb[index].duration = 0; /* Paranoia */ pc_cb[index].cb_func = NULL; /* Paranoia */ pc_cb[index].cb_active = 0; rw_exit(&pc_cb[index].cb_rwlock); /* * Just move all the entries to a private list which we can walk * through without the need to hold any locks. * No more requests can get added to the hash lists for this consumer * as the cb_active field for the callback has been cleared. */ for (i = 0; i < NUM_PAGE_CAPTURE_BUCKETS; i++) { mutex_enter(&page_capture_hash[i].pchh_mutex); for (j = 0; j < 2; j++) { bp1 = page_capture_hash[i].lists[j].next; /* walk through all but first (sentinel) element */ while (bp1 != &page_capture_hash[i].lists[j]) { bp2 = bp1; if (bp2->flags & flags) { bp1 = bp2->next; bp1->prev = bp2->prev; bp2->prev->next = bp1; bp2->next = head; head = bp2; /* * Clear the PR_CAPTURE bit as we * hold appropriate locks here. */ page_clrtoxic(head->pp, PR_CAPTURE); page_capture_hash[i]. num_pages[bp2->pri]--; continue; } bp1 = bp1->next; } } mutex_exit(&page_capture_hash[i].pchh_mutex); } while (head != NULL) { bp1 = head; head = head->next; kmem_free(bp1, sizeof (*bp1)); } } /* * Find pp in the active list and move it to the walked list if it * exists. * Note that most often pp should be at the front of the active list * as it is currently used and thus there is no other sort of optimization * being done here as this is a linked list data structure. * Returns 1 on successful move or 0 if page could not be found. */ static int page_capture_move_to_walked(page_t *pp) { page_capture_hash_bucket_t *bp; int index; index = PAGE_CAPTURE_HASH(pp); mutex_enter(&page_capture_hash[index].pchh_mutex); bp = page_capture_hash[index].lists[0].next; while (bp != &page_capture_hash[index].lists[0]) { if (bp->pp == pp) { /* Remove from old list */ bp->next->prev = bp->prev; bp->prev->next = bp->next; /* Add to new list */ bp->next = page_capture_hash[index].lists[1].next; bp->prev = &page_capture_hash[index].lists[1]; page_capture_hash[index].lists[1].next = bp; bp->next->prev = bp; /* * There is a small probability of page on a free * list being retired while being allocated * and before P_RAF is set on it. The page may * end up marked as high priority request instead * of low priority request. * If P_RAF page is not marked as low priority request * change it to low priority request. */ page_capture_hash[index].num_pages[bp->pri]--; bp->pri = PAGE_CAPTURE_PRIO(pp); page_capture_hash[index].num_pages[bp->pri]++; mutex_exit(&page_capture_hash[index].pchh_mutex); return (1); } bp = bp->next; } mutex_exit(&page_capture_hash[index].pchh_mutex); return (0); } /* * Add a new entry to the page capture hash. The only case where a new * entry is not added is when the page capture consumer is no longer registered. * In this case, we'll silently not add the page to the hash. We know that * page retire will always be registered for the case where we are currently * unretiring a page and thus there are no conflicts. */ static void page_capture_add_hash(page_t *pp, uint_t szc, uint_t flags, void *datap) { page_capture_hash_bucket_t *bp1; page_capture_hash_bucket_t *bp2; int index; int cb_index; int i; uchar_t pri; #ifdef DEBUG page_capture_hash_bucket_t *tp1; int l; #endif ASSERT(!(flags & CAPTURE_ASYNC)); bp1 = kmem_alloc(sizeof (struct page_capture_hash_bucket), KM_SLEEP); bp1->pp = pp; bp1->szc = szc; bp1->flags = flags; bp1->datap = datap; for (cb_index = 0; cb_index < PC_NUM_CALLBACKS; cb_index++) { if ((flags >> cb_index) & 1) { break; } } ASSERT(cb_index != PC_NUM_CALLBACKS); rw_enter(&pc_cb[cb_index].cb_rwlock, RW_READER); if (pc_cb[cb_index].cb_active) { if (pc_cb[cb_index].duration == -1) { bp1->expires = (clock_t)-1; } else { bp1->expires = ddi_get_lbolt() + pc_cb[cb_index].duration; } } else { /* There's no callback registered so don't add to the hash */ rw_exit(&pc_cb[cb_index].cb_rwlock); kmem_free(bp1, sizeof (*bp1)); return; } index = PAGE_CAPTURE_HASH(pp); /* * Only allow capture flag to be modified under this mutex. * Prevents multiple entries for same page getting added. */ mutex_enter(&page_capture_hash[index].pchh_mutex); /* * if not already on the hash, set capture bit and add to the hash */ if (!(pp->p_toxic & PR_CAPTURE)) { #ifdef DEBUG /* Check for duplicate entries */ for (l = 0; l < 2; l++) { tp1 = page_capture_hash[index].lists[l].next; while (tp1 != &page_capture_hash[index].lists[l]) { if (tp1->pp == pp) { panic("page pp 0x%p already on hash " "at 0x%p\n", (void *)pp, (void *)tp1); } tp1 = tp1->next; } } #endif page_settoxic(pp, PR_CAPTURE); pri = PAGE_CAPTURE_PRIO(pp); bp1->pri = pri; bp1->next = page_capture_hash[index].lists[0].next; bp1->prev = &page_capture_hash[index].lists[0]; bp1->next->prev = bp1; page_capture_hash[index].lists[0].next = bp1; page_capture_hash[index].num_pages[pri]++; if (flags & CAPTURE_RETIRE) { page_retire_incr_pend_count(datap); } mutex_exit(&page_capture_hash[index].pchh_mutex); rw_exit(&pc_cb[cb_index].cb_rwlock); cv_signal(&pc_cv); return; } /* * A page retire request will replace any other request. * A second physmem request which is for a different process than * the currently registered one will be dropped as there is * no way to hold the private data for both calls. * In the future, once there are more callers, this will have to * be worked out better as there needs to be private storage for * at least each type of caller (maybe have datap be an array of * *void's so that we can index based upon callers index). */ /* walk hash list to update expire time */ for (i = 0; i < 2; i++) { bp2 = page_capture_hash[index].lists[i].next; while (bp2 != &page_capture_hash[index].lists[i]) { if (bp2->pp == pp) { if (flags & CAPTURE_RETIRE) { if (!(bp2->flags & CAPTURE_RETIRE)) { page_retire_incr_pend_count( datap); bp2->flags = flags; bp2->expires = bp1->expires; bp2->datap = datap; } } else { ASSERT(flags & CAPTURE_PHYSMEM); if (!(bp2->flags & CAPTURE_RETIRE) && (datap == bp2->datap)) { bp2->expires = bp1->expires; } } mutex_exit(&page_capture_hash[index]. pchh_mutex); rw_exit(&pc_cb[cb_index].cb_rwlock); kmem_free(bp1, sizeof (*bp1)); return; } bp2 = bp2->next; } } /* * the PR_CAPTURE flag is protected by the page_capture_hash mutexes * and thus it either has to be set or not set and can't change * while holding the mutex above. */ panic("page_capture_add_hash, PR_CAPTURE flag set on pp %p\n", (void *)pp); } /* * We have a page in our hands, lets try and make it ours by turning * it into a clean page like it had just come off the freelists. * * Returns 0 on success, with the page still EXCL locked. * On failure, the page will be unlocked, and returns EAGAIN */ static int page_capture_clean_page(page_t *pp) { page_t *newpp; int skip_unlock = 0; spgcnt_t count; page_t *tpp; int ret = 0; int extra; ASSERT(PAGE_EXCL(pp)); ASSERT(!PP_RETIRED(pp)); ASSERT(curthread->t_flag & T_CAPTURING); if (PP_ISFREE(pp)) { if (!page_reclaim(pp, NULL)) { skip_unlock = 1; ret = EAGAIN; goto cleanup; } ASSERT(pp->p_szc == 0); if (pp->p_vnode != NULL) { /* * Since this page came from the * cachelist, we must destroy the * old vnode association. */ page_hashout(pp, NULL); } goto cleanup; } /* * If we know page_relocate will fail, skip it * It could still fail due to a UE on another page but we * can't do anything about that. */ if (pp->p_toxic & PR_UE) { goto skip_relocate; } /* * It's possible that pages can not have a vnode as fsflush comes * through and cleans up these pages. It's ugly but that's how it is. */ if (pp->p_vnode == NULL) { goto skip_relocate; } /* * Page was not free, so lets try to relocate it. * page_relocate only works with root pages, so if this is not a root * page, we need to demote it to try and relocate it. * Unfortunately this is the best we can do right now. */ newpp = NULL; if ((pp->p_szc > 0) && (pp != PP_PAGEROOT(pp))) { if (page_try_demote_pages(pp) == 0) { ret = EAGAIN; goto cleanup; } } ret = page_relocate(&pp, &newpp, 1, 0, &count, NULL); if (ret == 0) { page_t *npp; /* unlock the new page(s) */ while (count-- > 0) { ASSERT(newpp != NULL); npp = newpp; page_sub(&newpp, npp); page_unlock(npp); } ASSERT(newpp == NULL); /* * Check to see if the page we have is too large. * If so, demote it freeing up the extra pages. */ if (pp->p_szc > 0) { /* For now demote extra pages to szc == 0 */ extra = page_get_pagecnt(pp->p_szc) - 1; while (extra > 0) { tpp = pp->p_next; page_sub(&pp, tpp); tpp->p_szc = 0; page_free(tpp, 1); extra--; } /* Make sure to set our page to szc 0 as well */ ASSERT(pp->p_next == pp && pp->p_prev == pp); pp->p_szc = 0; } goto cleanup; } else if (ret == EIO) { ret = EAGAIN; goto cleanup; } else { /* * Need to reset return type as we failed to relocate the page * but that does not mean that some of the next steps will not * work. */ ret = 0; } skip_relocate: if (pp->p_szc > 0) { if (page_try_demote_pages(pp) == 0) { ret = EAGAIN; goto cleanup; } } ASSERT(pp->p_szc == 0); if (hat_ismod(pp)) { ret = EAGAIN; goto cleanup; } if (PP_ISKAS(pp)) { ret = EAGAIN; goto cleanup; } if (pp->p_lckcnt || pp->p_cowcnt) { ret = EAGAIN; goto cleanup; } (void) hat_pageunload(pp, HAT_FORCE_PGUNLOAD); ASSERT(!hat_page_is_mapped(pp)); if (hat_ismod(pp)) { /* * This is a semi-odd case as the page is now modified but not * mapped as we just unloaded the mappings above. */ ret = EAGAIN; goto cleanup; } if (pp->p_vnode != NULL) { page_hashout(pp, NULL); } /* * At this point, the page should be in a clean state and * we can do whatever we want with it. */ cleanup: if (ret != 0) { if (!skip_unlock) { page_unlock(pp); } } else { ASSERT(pp->p_szc == 0); ASSERT(PAGE_EXCL(pp)); pp->p_next = pp; pp->p_prev = pp; } return (ret); } /* * Various callers of page_trycapture() can have different restrictions upon * what memory they have access to. * Returns 0 on success, with the following error codes on failure: * EPERM - The requested page is long term locked, and thus repeated * requests to capture this page will likely fail. * ENOMEM - There was not enough free memory in the system to safely * map the requested page. * ENOENT - The requested page was inside the kernel cage, and the * PHYSMEM_CAGE flag was not set. */ int page_capture_pre_checks(page_t *pp, uint_t flags) { ASSERT(pp != NULL); #if defined(__sparc) if (pp->p_vnode == &promvp) { return (EPERM); } if (PP_ISNORELOC(pp) && !(flags & CAPTURE_GET_CAGE) && (flags & CAPTURE_PHYSMEM)) { return (ENOENT); } if (PP_ISNORELOCKERNEL(pp)) { return (EPERM); } #else if (PP_ISKAS(pp)) { return (EPERM); } #endif /* __sparc */ /* only physmem currently has the restrictions checked below */ if (!(flags & CAPTURE_PHYSMEM)) { return (0); } if (availrmem < swapfs_minfree) { /* * We won't try to capture this page as we are * running low on memory. */ return (ENOMEM); } return (0); } /* * Once we have a page in our mits, go ahead and complete the capture * operation. * Returns 1 on failure where page is no longer needed * Returns 0 on success * Returns -1 if there was a transient failure. * Failure cases must release the SE_EXCL lock on pp (usually via page_free). */ int page_capture_take_action(page_t *pp, uint_t flags, void *datap) { int cb_index; int ret = 0; page_capture_hash_bucket_t *bp1; page_capture_hash_bucket_t *bp2; int index; int found = 0; int i; ASSERT(PAGE_EXCL(pp)); ASSERT(curthread->t_flag & T_CAPTURING); for (cb_index = 0; cb_index < PC_NUM_CALLBACKS; cb_index++) { if ((flags >> cb_index) & 1) { break; } } ASSERT(cb_index < PC_NUM_CALLBACKS); /* * Remove the entry from the page_capture hash, but don't free it yet * as we may need to put it back. * Since we own the page at this point in time, we should find it * in the hash if this is an ASYNC call. If we don't it's likely * that the page_capture_async() thread decided that this request * had expired, in which case we just continue on. */ if (flags & CAPTURE_ASYNC) { index = PAGE_CAPTURE_HASH(pp); mutex_enter(&page_capture_hash[index].pchh_mutex); for (i = 0; i < 2 && !found; i++) { bp1 = page_capture_hash[index].lists[i].next; while (bp1 != &page_capture_hash[index].lists[i]) { if (bp1->pp == pp) { bp1->next->prev = bp1->prev; bp1->prev->next = bp1->next; page_capture_hash[index]. num_pages[bp1->pri]--; page_clrtoxic(pp, PR_CAPTURE); found = 1; break; } bp1 = bp1->next; } } mutex_exit(&page_capture_hash[index].pchh_mutex); } /* Synchronize with the unregister func. */ rw_enter(&pc_cb[cb_index].cb_rwlock, RW_READER); if (!pc_cb[cb_index].cb_active) { page_free(pp, 1); rw_exit(&pc_cb[cb_index].cb_rwlock); if (found) { kmem_free(bp1, sizeof (*bp1)); } return (1); } /* * We need to remove the entry from the page capture hash and turn off * the PR_CAPTURE bit before calling the callback. We'll need to cache * the entry here, and then based upon the return value, cleanup * appropriately or re-add it to the hash, making sure that someone else * hasn't already done so. * It should be rare for the callback to fail and thus it's ok for * the failure path to be a bit complicated as the success path is * cleaner and the locking rules are easier to follow. */ ret = pc_cb[cb_index].cb_func(pp, datap, flags); rw_exit(&pc_cb[cb_index].cb_rwlock); /* * If this was an ASYNC request, we need to cleanup the hash if the * callback was successful or if the request was no longer valid. * For non-ASYNC requests, we return failure to map and the caller * will take care of adding the request to the hash. * Note also that the callback itself is responsible for the page * at this point in time in terms of locking ... The most common * case for the failure path should just be a page_free. */ if (ret >= 0) { if (found) { if (bp1->flags & CAPTURE_RETIRE) { page_retire_decr_pend_count(datap); } kmem_free(bp1, sizeof (*bp1)); } return (ret); } if (!found) { return (ret); } ASSERT(flags & CAPTURE_ASYNC); /* * Check for expiration time first as we can just free it up if it's * expired. */ if (ddi_get_lbolt() > bp1->expires && bp1->expires != -1) { kmem_free(bp1, sizeof (*bp1)); return (ret); } /* * The callback failed and there used to be an entry in the hash for * this page, so we need to add it back to the hash. */ mutex_enter(&page_capture_hash[index].pchh_mutex); if (!(pp->p_toxic & PR_CAPTURE)) { /* just add bp1 back to head of walked list */ page_settoxic(pp, PR_CAPTURE); bp1->next = page_capture_hash[index].lists[1].next; bp1->prev = &page_capture_hash[index].lists[1]; bp1->next->prev = bp1; bp1->pri = PAGE_CAPTURE_PRIO(pp); page_capture_hash[index].lists[1].next = bp1; page_capture_hash[index].num_pages[bp1->pri]++; mutex_exit(&page_capture_hash[index].pchh_mutex); return (ret); } /* * Otherwise there was a new capture request added to list * Need to make sure that our original data is represented if * appropriate. */ for (i = 0; i < 2; i++) { bp2 = page_capture_hash[index].lists[i].next; while (bp2 != &page_capture_hash[index].lists[i]) { if (bp2->pp == pp) { if (bp1->flags & CAPTURE_RETIRE) { if (!(bp2->flags & CAPTURE_RETIRE)) { bp2->szc = bp1->szc; bp2->flags = bp1->flags; bp2->expires = bp1->expires; bp2->datap = bp1->datap; } } else { ASSERT(bp1->flags & CAPTURE_PHYSMEM); if (!(bp2->flags & CAPTURE_RETIRE)) { bp2->szc = bp1->szc; bp2->flags = bp1->flags; bp2->expires = bp1->expires; bp2->datap = bp1->datap; } } page_capture_hash[index].num_pages[bp2->pri]--; bp2->pri = PAGE_CAPTURE_PRIO(pp); page_capture_hash[index].num_pages[bp2->pri]++; mutex_exit(&page_capture_hash[index]. pchh_mutex); kmem_free(bp1, sizeof (*bp1)); return (ret); } bp2 = bp2->next; } } panic("PR_CAPTURE set but not on hash for pp 0x%p\n", (void *)pp); /*NOTREACHED*/ } /* * Try to capture the given page for the caller specified in the flags * parameter. The page will either be captured and handed over to the * appropriate callback, or will be queued up in the page capture hash * to be captured asynchronously. * If the current request is due to an async capture, the page must be * exclusively locked before calling this function. * Currently szc must be 0 but in the future this should be expandable to * other page sizes. * Returns 0 on success, with the following error codes on failure: * EPERM - The requested page is long term locked, and thus repeated * requests to capture this page will likely fail. * ENOMEM - There was not enough free memory in the system to safely * map the requested page. * ENOENT - The requested page was inside the kernel cage, and the * CAPTURE_GET_CAGE flag was not set. * EAGAIN - The requested page could not be capturead at this point in * time but future requests will likely work. * EBUSY - The requested page is retired and the CAPTURE_GET_RETIRED flag * was not set. */ int page_itrycapture(page_t *pp, uint_t szc, uint_t flags, void *datap) { int ret; int cb_index; if (flags & CAPTURE_ASYNC) { ASSERT(PAGE_EXCL(pp)); goto async; } /* Make sure there's enough availrmem ... */ ret = page_capture_pre_checks(pp, flags); if (ret != 0) { return (ret); } if (!page_trylock(pp, SE_EXCL)) { for (cb_index = 0; cb_index < PC_NUM_CALLBACKS; cb_index++) { if ((flags >> cb_index) & 1) { break; } } ASSERT(cb_index < PC_NUM_CALLBACKS); ret = EAGAIN; /* Special case for retired pages */ if (PP_RETIRED(pp)) { if (flags & CAPTURE_GET_RETIRED) { if (!page_unretire_pp(pp, PR_UNR_TEMP)) { /* * Need to set capture bit and add to * hash so that the page will be * retired when freed. */ page_capture_add_hash(pp, szc, CAPTURE_RETIRE, NULL); ret = 0; goto own_page; } } else { return (EBUSY); } } page_capture_add_hash(pp, szc, flags, datap); return (ret); } async: ASSERT(PAGE_EXCL(pp)); /* Need to check for physmem async requests that availrmem is sane */ if ((flags & (CAPTURE_ASYNC | CAPTURE_PHYSMEM)) == (CAPTURE_ASYNC | CAPTURE_PHYSMEM) && (availrmem < swapfs_minfree)) { page_unlock(pp); return (ENOMEM); } ret = page_capture_clean_page(pp); if (ret != 0) { /* We failed to get the page, so lets add it to the hash */ if (!(flags & CAPTURE_ASYNC)) { page_capture_add_hash(pp, szc, flags, datap); } return (ret); } own_page: ASSERT(PAGE_EXCL(pp)); ASSERT(pp->p_szc == 0); /* Call the callback */ ret = page_capture_take_action(pp, flags, datap); if (ret == 0) { return (0); } /* * Note that in the failure cases from page_capture_take_action, the * EXCL lock will have already been dropped. */ if ((ret == -1) && (!(flags & CAPTURE_ASYNC))) { page_capture_add_hash(pp, szc, flags, datap); } return (EAGAIN); } int page_trycapture(page_t *pp, uint_t szc, uint_t flags, void *datap) { int ret; curthread->t_flag |= T_CAPTURING; ret = page_itrycapture(pp, szc, flags, datap); curthread->t_flag &= ~T_CAPTURING; /* xor works as we know its set */ return (ret); } /* * When unlocking a page which has the PR_CAPTURE bit set, this routine * gets called to try and capture the page. */ void page_unlock_capture(page_t *pp) { page_capture_hash_bucket_t *bp; int index; int i; uint_t szc; uint_t flags = 0; void *datap; kmutex_t *mp; extern vnode_t retired_pages; /* * We need to protect against a possible deadlock here where we own * the vnode page hash mutex and want to acquire it again as there * are locations in the code, where we unlock a page while holding * the mutex which can lead to the page being captured and eventually * end up here. As we may be hashing out the old page and hashing into * the retire vnode, we need to make sure we don't own them. * Other callbacks who do hash operations also need to make sure that * before they hashin to a vnode that they do not currently own the * vphm mutex otherwise there will be a panic. */ if (mutex_owned(page_vnode_mutex(&retired_pages))) { page_unlock_nocapture(pp); return; } if (pp->p_vnode != NULL && mutex_owned(page_vnode_mutex(pp->p_vnode))) { page_unlock_nocapture(pp); return; } index = PAGE_CAPTURE_HASH(pp); mp = &page_capture_hash[index].pchh_mutex; mutex_enter(mp); for (i = 0; i < 2; i++) { bp = page_capture_hash[index].lists[i].next; while (bp != &page_capture_hash[index].lists[i]) { if (bp->pp == pp) { szc = bp->szc; flags = bp->flags | CAPTURE_ASYNC; datap = bp->datap; mutex_exit(mp); (void) page_trycapture(pp, szc, flags, datap); return; } bp = bp->next; } } /* Failed to find page in hash so clear flags and unlock it. */ page_clrtoxic(pp, PR_CAPTURE); page_unlock(pp); mutex_exit(mp); } void page_capture_init() { int i; for (i = 0; i < NUM_PAGE_CAPTURE_BUCKETS; i++) { page_capture_hash[i].lists[0].next = &page_capture_hash[i].lists[0]; page_capture_hash[i].lists[0].prev = &page_capture_hash[i].lists[0]; page_capture_hash[i].lists[1].next = &page_capture_hash[i].lists[1]; page_capture_hash[i].lists[1].prev = &page_capture_hash[i].lists[1]; } pc_thread_shortwait = 23 * hz; pc_thread_longwait = 1201 * hz; pc_thread_retry = 3; mutex_init(&pc_thread_mutex, NULL, MUTEX_DEFAULT, NULL); cv_init(&pc_cv, NULL, CV_DEFAULT, NULL); pc_thread_id = thread_create(NULL, 0, page_capture_thread, NULL, 0, &p0, TS_RUN, minclsyspri); } /* * It is necessary to scrub any failing pages prior to reboot in order to * prevent a latent error trap from occurring on the next boot. */ void page_retire_mdboot() { page_t *pp; int i, j; page_capture_hash_bucket_t *bp; uchar_t pri; /* walk lists looking for pages to scrub */ for (i = 0; i < NUM_PAGE_CAPTURE_BUCKETS; i++) { for (pri = 0; pri < PC_NUM_PRI; pri++) { if (page_capture_hash[i].num_pages[pri] != 0) { break; } } if (pri == PC_NUM_PRI) continue; mutex_enter(&page_capture_hash[i].pchh_mutex); for (j = 0; j < 2; j++) { bp = page_capture_hash[i].lists[j].next; while (bp != &page_capture_hash[i].lists[j]) { pp = bp->pp; if (PP_TOXIC(pp)) { if (page_trylock(pp, SE_EXCL)) { PP_CLRFREE(pp); pagescrub(pp, 0, PAGESIZE); page_unlock(pp); } } bp = bp->next; } } mutex_exit(&page_capture_hash[i].pchh_mutex); } } /* * Walk the page_capture_hash trying to capture pages and also cleanup old * entries which have expired. */ void page_capture_async() { page_t *pp; int i; int ret; page_capture_hash_bucket_t *bp1, *bp2; uint_t szc; uint_t flags; void *datap; uchar_t pri; /* If there are outstanding pages to be captured, get to work */ for (i = 0; i < NUM_PAGE_CAPTURE_BUCKETS; i++) { for (pri = 0; pri < PC_NUM_PRI; pri++) { if (page_capture_hash[i].num_pages[pri] != 0) break; } if (pri == PC_NUM_PRI) continue; /* Append list 1 to list 0 and then walk through list 0 */ mutex_enter(&page_capture_hash[i].pchh_mutex); bp1 = &page_capture_hash[i].lists[1]; bp2 = bp1->next; if (bp1 != bp2) { bp1->prev->next = page_capture_hash[i].lists[0].next; bp2->prev = &page_capture_hash[i].lists[0]; page_capture_hash[i].lists[0].next->prev = bp1->prev; page_capture_hash[i].lists[0].next = bp2; bp1->next = bp1; bp1->prev = bp1; } /* list[1] will be empty now */ bp1 = page_capture_hash[i].lists[0].next; while (bp1 != &page_capture_hash[i].lists[0]) { /* Check expiration time */ if ((ddi_get_lbolt() > bp1->expires && bp1->expires != -1) || page_deleted(bp1->pp)) { page_capture_hash[i].lists[0].next = bp1->next; bp1->next->prev = &page_capture_hash[i].lists[0]; page_capture_hash[i].num_pages[bp1->pri]--; /* * We can safely remove the PR_CAPTURE bit * without holding the EXCL lock on the page * as the PR_CAPTURE bit requres that the * page_capture_hash[].pchh_mutex be held * to modify it. */ page_clrtoxic(bp1->pp, PR_CAPTURE); mutex_exit(&page_capture_hash[i].pchh_mutex); kmem_free(bp1, sizeof (*bp1)); mutex_enter(&page_capture_hash[i].pchh_mutex); bp1 = page_capture_hash[i].lists[0].next; continue; } pp = bp1->pp; szc = bp1->szc; flags = bp1->flags; datap = bp1->datap; mutex_exit(&page_capture_hash[i].pchh_mutex); if (page_trylock(pp, SE_EXCL)) { ret = page_trycapture(pp, szc, flags | CAPTURE_ASYNC, datap); } else { ret = 1; /* move to walked hash */ } if (ret != 0) { /* Move to walked hash */ (void) page_capture_move_to_walked(pp); } mutex_enter(&page_capture_hash[i].pchh_mutex); bp1 = page_capture_hash[i].lists[0].next; } mutex_exit(&page_capture_hash[i].pchh_mutex); } } /* * This function is called by the page_capture_thread, and is needed in * in order to initiate aio cleanup, so that pages used in aio * will be unlocked and subsequently retired by page_capture_thread. */ static int do_aio_cleanup(void) { proc_t *procp; int (*aio_cleanup_dr_delete_memory)(proc_t *); int cleaned = 0; if (modload("sys", "kaio") == -1) { cmn_err(CE_WARN, "do_aio_cleanup: cannot load kaio"); return (0); } /* * We use the aio_cleanup_dr_delete_memory function to * initiate the actual clean up; this function will wake * up the per-process aio_cleanup_thread. */ aio_cleanup_dr_delete_memory = (int (*)(proc_t *)) modgetsymvalue("aio_cleanup_dr_delete_memory", 0); if (aio_cleanup_dr_delete_memory == NULL) { cmn_err(CE_WARN, "aio_cleanup_dr_delete_memory not found in kaio"); return (0); } mutex_enter(&pidlock); for (procp = practive; (procp != NULL); procp = procp->p_next) { mutex_enter(&procp->p_lock); if (procp->p_aio != NULL) { /* cleanup proc's outstanding kaio */ cleaned += (*aio_cleanup_dr_delete_memory)(procp); } mutex_exit(&procp->p_lock); } mutex_exit(&pidlock); return (cleaned); } /* * helper function for page_capture_thread */ static void page_capture_handle_outstanding(void) { int ntry; /* Reap pages before attempting capture pages */ kmem_reap(); if ((page_retire_pend_count() > page_retire_pend_kas_count()) && hat_supported(HAT_DYNAMIC_ISM_UNMAP, (void *)0)) { /* * Note: Purging only for platforms that support * ISM hat_pageunload() - mainly SPARC. On x86/x64 * platforms ISM pages SE_SHARED locked until destroyed. */ /* disable and purge seg_pcache */ (void) seg_p_disable(); for (ntry = 0; ntry < pc_thread_retry; ntry++) { if (!page_retire_pend_count()) break; if (do_aio_cleanup()) { /* * allow the apps cleanup threads * to run */ delay(pc_thread_shortwait); } page_capture_async(); } /* reenable seg_pcache */ seg_p_enable(); /* completed what can be done. break out */ return; } /* * For kernel pages and/or unsupported HAT_DYNAMIC_ISM_UNMAP, reap * and then attempt to capture. */ seg_preap(); page_capture_async(); } /* * The page_capture_thread loops forever, looking to see if there are * pages still waiting to be captured. */ static void page_capture_thread(void) { callb_cpr_t c; int i; int high_pri_pages; int low_pri_pages; clock_t timeout; CALLB_CPR_INIT(&c, &pc_thread_mutex, callb_generic_cpr, "page_capture"); mutex_enter(&pc_thread_mutex); for (;;) { high_pri_pages = 0; low_pri_pages = 0; for (i = 0; i < NUM_PAGE_CAPTURE_BUCKETS; i++) { high_pri_pages += page_capture_hash[i].num_pages[PC_PRI_HI]; low_pri_pages += page_capture_hash[i].num_pages[PC_PRI_LO]; } timeout = pc_thread_longwait; if (high_pri_pages != 0) { timeout = pc_thread_shortwait; page_capture_handle_outstanding(); } else if (low_pri_pages != 0) { page_capture_async(); } CALLB_CPR_SAFE_BEGIN(&c); (void) cv_reltimedwait(&pc_cv, &pc_thread_mutex, timeout, TR_CLOCK_TICK); CALLB_CPR_SAFE_END(&c, &pc_thread_mutex); } /*NOTREACHED*/ } /* * Attempt to locate a bucket that has enough pages to satisfy the request. * The initial check is done without the lock to avoid unneeded contention. * The function returns 1 if enough pages were found, else 0 if it could not * find enough pages in a bucket. */ static int pcf_decrement_bucket(pgcnt_t npages) { struct pcf *p; struct pcf *q; int i; p = &pcf[PCF_INDEX()]; q = &pcf[pcf_fanout]; for (i = 0; i < pcf_fanout; i++) { if (p->pcf_count > npages) { /* * a good one to try. */ mutex_enter(&p->pcf_lock); if (p->pcf_count > npages) { p->pcf_count -= (uint_t)npages; /* * freemem is not protected by any lock. * Thus, we cannot have any assertion * containing freemem here. */ freemem -= npages; mutex_exit(&p->pcf_lock); return (1); } mutex_exit(&p->pcf_lock); } p++; if (p >= q) { p = pcf; } } return (0); } /* * Arguments: * pcftotal_ret: If the value is not NULL and we have walked all the * buckets but did not find enough pages then it will * be set to the total number of pages in all the pcf * buckets. * npages: Is the number of pages we have been requested to * find. * unlock: If set to 0 we will leave the buckets locked if the * requested number of pages are not found. * * Go and try to satisfy the page request from any number of buckets. * This can be a very expensive operation as we have to lock the buckets * we are checking (and keep them locked), starting at bucket 0. * * The function returns 1 if enough pages were found, else 0 if it could not * find enough pages in the buckets. * */ static int pcf_decrement_multiple(pgcnt_t *pcftotal_ret, pgcnt_t npages, int unlock) { struct pcf *p; pgcnt_t pcftotal; int i; p = pcf; /* try to collect pages from several pcf bins */ for (pcftotal = 0, i = 0; i < pcf_fanout; i++) { mutex_enter(&p->pcf_lock); pcftotal += p->pcf_count; if (pcftotal >= npages) { /* * Wow! There are enough pages laying around * to satisfy the request. Do the accounting, * drop the locks we acquired, and go back. * * freemem is not protected by any lock. So, * we cannot have any assertion containing * freemem. */ freemem -= npages; while (p >= pcf) { if (p->pcf_count <= npages) { npages -= p->pcf_count; p->pcf_count = 0; } else { p->pcf_count -= (uint_t)npages; npages = 0; } mutex_exit(&p->pcf_lock); p--; } ASSERT(npages == 0); return (1); } p++; } if (unlock) { /* failed to collect pages - release the locks */ while (--p >= pcf) { mutex_exit(&p->pcf_lock); } } if (pcftotal_ret != NULL) *pcftotal_ret = pcftotal; return (0); }