1e2068d0bSJeff Roberson /*- 2e2068d0bSJeff Roberson * SPDX-License-Identifier: (BSD-3-Clause AND MIT-CMU) 3e2068d0bSJeff Roberson * 4e2068d0bSJeff Roberson * Copyright (c) 1991, 1993 5e2068d0bSJeff Roberson * The Regents of the University of California. All rights reserved. 6e2068d0bSJeff Roberson * 7e2068d0bSJeff Roberson * This code is derived from software contributed to Berkeley by 8e2068d0bSJeff Roberson * The Mach Operating System project at Carnegie-Mellon University. 9e2068d0bSJeff Roberson * 10e2068d0bSJeff Roberson * Redistribution and use in source and binary forms, with or without 11e2068d0bSJeff Roberson * modification, are permitted provided that the following conditions 12e2068d0bSJeff Roberson * are met: 13e2068d0bSJeff Roberson * 1. Redistributions of source code must retain the above copyright 14e2068d0bSJeff Roberson * notice, this list of conditions and the following disclaimer. 15e2068d0bSJeff Roberson * 2. Redistributions in binary form must reproduce the above copyright 16e2068d0bSJeff Roberson * notice, this list of conditions and the following disclaimer in the 17e2068d0bSJeff Roberson * documentation and/or other materials provided with the distribution. 18e2068d0bSJeff Roberson * 3. Neither the name of the University nor the names of its contributors 19e2068d0bSJeff Roberson * may be used to endorse or promote products derived from this software 20e2068d0bSJeff Roberson * without specific prior written permission. 21e2068d0bSJeff Roberson * 22e2068d0bSJeff Roberson * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23e2068d0bSJeff Roberson * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24e2068d0bSJeff Roberson * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25e2068d0bSJeff Roberson * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26e2068d0bSJeff Roberson * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27e2068d0bSJeff Roberson * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28e2068d0bSJeff Roberson * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29e2068d0bSJeff Roberson * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30e2068d0bSJeff Roberson * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31e2068d0bSJeff Roberson * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32e2068d0bSJeff Roberson * SUCH DAMAGE. 33e2068d0bSJeff Roberson * 34e2068d0bSJeff Roberson * from: @(#)vm_page.h 8.2 (Berkeley) 12/13/93 35e2068d0bSJeff Roberson * 36e2068d0bSJeff Roberson * 37e2068d0bSJeff Roberson * Copyright (c) 1987, 1990 Carnegie-Mellon University. 38e2068d0bSJeff Roberson * All rights reserved. 39e2068d0bSJeff Roberson * 40e2068d0bSJeff Roberson * Authors: Avadis Tevanian, Jr., Michael Wayne Young 41e2068d0bSJeff Roberson * 42e2068d0bSJeff Roberson * Permission to use, copy, modify and distribute this software and 43e2068d0bSJeff Roberson * its documentation is hereby granted, provided that both the copyright 44e2068d0bSJeff Roberson * notice and this permission notice appear in all copies of the 45e2068d0bSJeff Roberson * software, derivative works or modified versions, and any portions 46e2068d0bSJeff Roberson * thereof, and that both notices appear in supporting documentation. 47e2068d0bSJeff Roberson * 48e2068d0bSJeff Roberson * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" 49e2068d0bSJeff Roberson * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND 50e2068d0bSJeff Roberson * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. 51e2068d0bSJeff Roberson * 52e2068d0bSJeff Roberson * Carnegie Mellon requests users of this software to return to 53e2068d0bSJeff Roberson * 54e2068d0bSJeff Roberson * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU 55e2068d0bSJeff Roberson * School of Computer Science 56e2068d0bSJeff Roberson * Carnegie Mellon University 57e2068d0bSJeff Roberson * Pittsburgh PA 15213-3890 58e2068d0bSJeff Roberson * 59e2068d0bSJeff Roberson * any improvements or extensions that they make and grant Carnegie the 60e2068d0bSJeff Roberson * rights to redistribute these changes. 61e2068d0bSJeff Roberson * 62e2068d0bSJeff Roberson * $FreeBSD$ 63e2068d0bSJeff Roberson */ 64e2068d0bSJeff Roberson 65e2068d0bSJeff Roberson #ifndef _VM_PAGEQUEUE_ 66e2068d0bSJeff Roberson #define _VM_PAGEQUEUE_ 67e2068d0bSJeff Roberson 68e2068d0bSJeff Roberson #ifdef _KERNEL 69e2068d0bSJeff Roberson struct vm_pagequeue { 70e2068d0bSJeff Roberson struct mtx pq_mutex; 71e2068d0bSJeff Roberson struct pglist pq_pl; 72e2068d0bSJeff Roberson int pq_cnt; 73e2068d0bSJeff Roberson const char * const pq_name; 74e2068d0bSJeff Roberson } __aligned(CACHE_LINE_SIZE); 75e2068d0bSJeff Roberson 765f8cd1c0SJeff Roberson #include <sys/pidctrl.h> 775f8cd1c0SJeff Roberson struct sysctl_oid; 78e2068d0bSJeff Roberson 7930fbfddaSJeff Roberson /* 8030fbfddaSJeff Roberson * One vm_domain per-numa domain. Contains pagequeues, free page structures, 8130fbfddaSJeff Roberson * and accounting. 8230fbfddaSJeff Roberson * 8330fbfddaSJeff Roberson * Lock Key: 8430fbfddaSJeff Roberson * f vmd_free_mtx 8530fbfddaSJeff Roberson * p vmd_pageout_mtx 8630fbfddaSJeff Roberson * d vm_domainset_lock 8730fbfddaSJeff Roberson * a atomic 8830fbfddaSJeff Roberson * c const after boot 89*60684862SMark Johnston * q page queue lock 9030fbfddaSJeff Roberson */ 91e2068d0bSJeff Roberson struct vm_domain { 92e2068d0bSJeff Roberson struct vm_pagequeue vmd_pagequeues[PQ_COUNT]; 93e2068d0bSJeff Roberson struct mtx_padalign vmd_free_mtx; 9430fbfddaSJeff Roberson struct mtx_padalign vmd_pageout_mtx; 9530fbfddaSJeff Roberson struct vmem *vmd_kernel_arena; /* (c) per-domain kva arena. */ 9630fbfddaSJeff Roberson u_int vmd_domain; /* (c) Domain number. */ 9730fbfddaSJeff Roberson u_int vmd_page_count; /* (c) Total page count. */ 9830fbfddaSJeff Roberson long vmd_segs; /* (c) bitmask of the segments */ 9930fbfddaSJeff Roberson u_int __aligned(CACHE_LINE_SIZE) vmd_free_count; /* (a,f) free page count */ 10030fbfddaSJeff Roberson u_int vmd_pageout_deficit; /* (a) Estimated number of pages deficit */ 10130fbfddaSJeff Roberson uint8_t vmd_pad[CACHE_LINE_SIZE - (sizeof(u_int) * 2)]; 102e2068d0bSJeff Roberson 10330fbfddaSJeff Roberson /* Paging control variables, used within single threaded page daemon. */ 1045f8cd1c0SJeff Roberson struct pidctrl vmd_pid; /* Pageout controller. */ 105e2068d0bSJeff Roberson boolean_t vmd_oom; 106e2068d0bSJeff Roberson int vmd_oom_seq; 107e2068d0bSJeff Roberson int vmd_last_active_scan; 108e2068d0bSJeff Roberson struct vm_page vmd_laundry_marker; 109e2068d0bSJeff Roberson struct vm_page vmd_marker; /* marker for pagedaemon private use */ 110e2068d0bSJeff Roberson struct vm_page vmd_inacthead; /* marker for LRU-defeating insertions */ 111e2068d0bSJeff Roberson 11230fbfddaSJeff Roberson int vmd_pageout_wanted; /* (a, p) pageout daemon wait channel */ 11330fbfddaSJeff Roberson int vmd_pageout_pages_needed; /* (d) page daemon waiting for pages? */ 11430fbfddaSJeff Roberson bool vmd_minset; /* (d) Are we in vm_min_domains? */ 11530fbfddaSJeff Roberson bool vmd_severeset; /* (d) Are we in vm_severe_domains? */ 116e2068d0bSJeff Roberson enum { 117e2068d0bSJeff Roberson VM_LAUNDRY_IDLE = 0, 118e2068d0bSJeff Roberson VM_LAUNDRY_BACKGROUND, 119e2068d0bSJeff Roberson VM_LAUNDRY_SHORTFALL 120e2068d0bSJeff Roberson } vmd_laundry_request; 121e2068d0bSJeff Roberson 122*60684862SMark Johnston /* Paging thresholds and targets. */ 123*60684862SMark Johnston u_int vmd_clean_pages_freed; /* (q) accumulator for laundry thread */ 124*60684862SMark Johnston u_int vmd_background_launder_target; /* (c) */ 125e2068d0bSJeff Roberson u_int vmd_free_reserved; /* (c) pages reserved for deadlock */ 126e2068d0bSJeff Roberson u_int vmd_free_target; /* (c) pages desired free */ 127e2068d0bSJeff Roberson u_int vmd_free_min; /* (c) pages desired free */ 128e2068d0bSJeff Roberson u_int vmd_inactive_target; /* (c) pages desired inactive */ 129e2068d0bSJeff Roberson u_int vmd_pageout_free_min; /* (c) min pages reserved for kernel */ 130e2068d0bSJeff Roberson u_int vmd_pageout_wakeup_thresh;/* (c) min pages to wake pagedaemon */ 131e2068d0bSJeff Roberson u_int vmd_interrupt_free_min; /* (c) reserved pages for int code */ 132e2068d0bSJeff Roberson u_int vmd_free_severe; /* (c) severe page depletion point */ 1335f8cd1c0SJeff Roberson 1345f8cd1c0SJeff Roberson /* Name for sysctl etc. */ 1355f8cd1c0SJeff Roberson struct sysctl_oid *vmd_oid; 1365f8cd1c0SJeff Roberson char vmd_name[sizeof(__XSTRING(MAXMEMDOM))]; 137e2068d0bSJeff Roberson } __aligned(CACHE_LINE_SIZE); 138e2068d0bSJeff Roberson 139e2068d0bSJeff Roberson extern struct vm_domain vm_dom[MAXMEMDOM]; 140e2068d0bSJeff Roberson 141e2068d0bSJeff Roberson #define VM_DOMAIN(n) (&vm_dom[(n)]) 142e2068d0bSJeff Roberson 143e2068d0bSJeff Roberson #define vm_pagequeue_assert_locked(pq) mtx_assert(&(pq)->pq_mutex, MA_OWNED) 144e2068d0bSJeff Roberson #define vm_pagequeue_lock(pq) mtx_lock(&(pq)->pq_mutex) 145e2068d0bSJeff Roberson #define vm_pagequeue_lockptr(pq) (&(pq)->pq_mutex) 146e2068d0bSJeff Roberson #define vm_pagequeue_unlock(pq) mtx_unlock(&(pq)->pq_mutex) 147e2068d0bSJeff Roberson 148e2068d0bSJeff Roberson #define vm_domain_free_assert_locked(n) \ 149e2068d0bSJeff Roberson mtx_assert(vm_domain_free_lockptr((n)), MA_OWNED) 150e2068d0bSJeff Roberson #define vm_domain_free_assert_unlocked(n) \ 151e2068d0bSJeff Roberson mtx_assert(vm_domain_free_lockptr((n)), MA_NOTOWNED) 152e2068d0bSJeff Roberson #define vm_domain_free_lock(d) \ 153e2068d0bSJeff Roberson mtx_lock(vm_domain_free_lockptr((d))) 154e2068d0bSJeff Roberson #define vm_domain_free_lockptr(d) \ 155e2068d0bSJeff Roberson (&(d)->vmd_free_mtx) 156e2068d0bSJeff Roberson #define vm_domain_free_unlock(d) \ 157e2068d0bSJeff Roberson mtx_unlock(vm_domain_free_lockptr((d))) 158e2068d0bSJeff Roberson 15930fbfddaSJeff Roberson #define vm_domain_pageout_lockptr(d) \ 16030fbfddaSJeff Roberson (&(d)->vmd_pageout_mtx) 16130fbfddaSJeff Roberson #define vm_domain_pageout_assert_locked(n) \ 16230fbfddaSJeff Roberson mtx_assert(vm_domain_pageout_lockptr((n)), MA_OWNED) 16330fbfddaSJeff Roberson #define vm_domain_pageout_assert_unlocked(n) \ 16430fbfddaSJeff Roberson mtx_assert(vm_domain_pageout_lockptr((n)), MA_NOTOWNED) 16530fbfddaSJeff Roberson #define vm_domain_pageout_lock(d) \ 16630fbfddaSJeff Roberson mtx_lock(vm_domain_pageout_lockptr((d))) 16730fbfddaSJeff Roberson #define vm_domain_pageout_unlock(d) \ 16830fbfddaSJeff Roberson mtx_unlock(vm_domain_pageout_lockptr((d))) 16930fbfddaSJeff Roberson 170e2068d0bSJeff Roberson static __inline void 171e2068d0bSJeff Roberson vm_pagequeue_cnt_add(struct vm_pagequeue *pq, int addend) 172e2068d0bSJeff Roberson { 173e2068d0bSJeff Roberson 174e2068d0bSJeff Roberson #ifdef notyet 175e2068d0bSJeff Roberson vm_pagequeue_assert_locked(pq); 176e2068d0bSJeff Roberson #endif 177e2068d0bSJeff Roberson pq->pq_cnt += addend; 178e2068d0bSJeff Roberson } 179e2068d0bSJeff Roberson #define vm_pagequeue_cnt_inc(pq) vm_pagequeue_cnt_add((pq), 1) 180e2068d0bSJeff Roberson #define vm_pagequeue_cnt_dec(pq) vm_pagequeue_cnt_add((pq), -1) 181e2068d0bSJeff Roberson 182e2068d0bSJeff Roberson void vm_domain_set(struct vm_domain *vmd); 18330fbfddaSJeff Roberson void vm_domain_clear(struct vm_domain *vmd); 1845c930c89SJeff Roberson int vm_domain_allocate(struct vm_domain *vmd, int req, int npages); 185e2068d0bSJeff Roberson 186e2068d0bSJeff Roberson /* 187e2068d0bSJeff Roberson * vm_pagequeue_domain: 188e2068d0bSJeff Roberson * 189e2068d0bSJeff Roberson * Return the memory domain the page belongs to. 190e2068d0bSJeff Roberson */ 191e2068d0bSJeff Roberson static inline struct vm_domain * 192e2068d0bSJeff Roberson vm_pagequeue_domain(vm_page_t m) 193e2068d0bSJeff Roberson { 194e2068d0bSJeff Roberson 195e2068d0bSJeff Roberson return (VM_DOMAIN(vm_phys_domain(m))); 196e2068d0bSJeff Roberson } 197e2068d0bSJeff Roberson 198e2068d0bSJeff Roberson /* 199e2068d0bSJeff Roberson * Return the number of pages we need to free-up or cache 200e2068d0bSJeff Roberson * A positive number indicates that we do not have enough free pages. 201e2068d0bSJeff Roberson */ 202e2068d0bSJeff Roberson static inline int 203e2068d0bSJeff Roberson vm_paging_target(struct vm_domain *vmd) 204e2068d0bSJeff Roberson { 205e2068d0bSJeff Roberson 206e2068d0bSJeff Roberson return (vmd->vmd_free_target - vmd->vmd_free_count); 207e2068d0bSJeff Roberson } 208e2068d0bSJeff Roberson 209e2068d0bSJeff Roberson /* 210e2068d0bSJeff Roberson * Returns TRUE if the pagedaemon needs to be woken up. 211e2068d0bSJeff Roberson */ 212e2068d0bSJeff Roberson static inline int 213e2068d0bSJeff Roberson vm_paging_needed(struct vm_domain *vmd, u_int free_count) 214e2068d0bSJeff Roberson { 215e2068d0bSJeff Roberson 216e2068d0bSJeff Roberson return (free_count < vmd->vmd_pageout_wakeup_thresh); 217e2068d0bSJeff Roberson } 218e2068d0bSJeff Roberson 219e2068d0bSJeff Roberson /* 220e2068d0bSJeff Roberson * Returns TRUE if the domain is below the min paging target. 221e2068d0bSJeff Roberson */ 222e2068d0bSJeff Roberson static inline int 223e2068d0bSJeff Roberson vm_paging_min(struct vm_domain *vmd) 224e2068d0bSJeff Roberson { 225e2068d0bSJeff Roberson 226e2068d0bSJeff Roberson return (vmd->vmd_free_min > vmd->vmd_free_count); 227e2068d0bSJeff Roberson } 228e2068d0bSJeff Roberson 229e2068d0bSJeff Roberson /* 230e2068d0bSJeff Roberson * Returns TRUE if the domain is below the severe paging target. 231e2068d0bSJeff Roberson */ 232e2068d0bSJeff Roberson static inline int 233e2068d0bSJeff Roberson vm_paging_severe(struct vm_domain *vmd) 234e2068d0bSJeff Roberson { 235e2068d0bSJeff Roberson 236e2068d0bSJeff Roberson return (vmd->vmd_free_severe > vmd->vmd_free_count); 237e2068d0bSJeff Roberson } 238e2068d0bSJeff Roberson 239e2068d0bSJeff Roberson /* 240e2068d0bSJeff Roberson * Return the number of pages we need to launder. 241e2068d0bSJeff Roberson * A positive number indicates that we have a shortfall of clean pages. 242e2068d0bSJeff Roberson */ 243e2068d0bSJeff Roberson static inline int 244e2068d0bSJeff Roberson vm_laundry_target(struct vm_domain *vmd) 245e2068d0bSJeff Roberson { 246e2068d0bSJeff Roberson 247e2068d0bSJeff Roberson return (vm_paging_target(vmd)); 248e2068d0bSJeff Roberson } 249e2068d0bSJeff Roberson 25030fbfddaSJeff Roberson void pagedaemon_wakeup(int domain); 25130fbfddaSJeff Roberson 25230fbfddaSJeff Roberson static inline void 25330fbfddaSJeff Roberson vm_domain_freecnt_inc(struct vm_domain *vmd, int adj) 254e2068d0bSJeff Roberson { 25530fbfddaSJeff Roberson u_int old, new; 256e2068d0bSJeff Roberson 25730fbfddaSJeff Roberson old = atomic_fetchadd_int(&vmd->vmd_free_count, adj); 25830fbfddaSJeff Roberson new = old + adj; 25930fbfddaSJeff Roberson /* 26030fbfddaSJeff Roberson * Only update bitsets on transitions. Notice we short-circuit the 26130fbfddaSJeff Roberson * rest of the checks if we're above min already. 26230fbfddaSJeff Roberson */ 26330fbfddaSJeff Roberson if (old < vmd->vmd_free_min && (new >= vmd->vmd_free_min || 26430fbfddaSJeff Roberson (old < vmd->vmd_free_severe && new >= vmd->vmd_free_severe) || 26530fbfddaSJeff Roberson (old < vmd->vmd_pageout_free_min && 26630fbfddaSJeff Roberson new >= vmd->vmd_pageout_free_min))) 26730fbfddaSJeff Roberson vm_domain_clear(vmd); 26830fbfddaSJeff Roberson } 26930fbfddaSJeff Roberson 270e2068d0bSJeff Roberson #endif /* _KERNEL */ 271e2068d0bSJeff Roberson #endif /* !_VM_PAGEQUEUE_ */ 272