From owner-svn-src-stable@FreeBSD.ORG Thu Jul 24 16:29:46 2014 Return-Path: Delivered-To: svn-src-stable@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) (using TLSv1 with cipher ADH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTPS id 4BDEC79C; Thu, 24 Jul 2014 16:29:46 +0000 (UTC) Received: from svn.freebsd.org (svn.freebsd.org [IPv6:2001:1900:2254:2068::e6a:0]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client did not present a certificate) by mx1.freebsd.org (Postfix) with ESMTPS id 3733321D7; Thu, 24 Jul 2014 16:29:46 +0000 (UTC) Received: from svn.freebsd.org ([127.0.1.70]) by svn.freebsd.org (8.14.8/8.14.8) with ESMTP id s6OGTk32029332; Thu, 24 Jul 2014 16:29:46 GMT (envelope-from kib@svn.freebsd.org) Received: (from kib@localhost) by svn.freebsd.org (8.14.8/8.14.8/Submit) id s6OGTi4e029324; Thu, 24 Jul 2014 16:29:44 GMT (envelope-from kib@svn.freebsd.org) Message-Id: <201407241629.s6OGTi4e029324@svn.freebsd.org> From: Konstantin Belousov Date: Thu, 24 Jul 2014 16:29:44 +0000 (UTC) To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-stable@freebsd.org, svn-src-stable-10@freebsd.org Subject: svn commit: r269072 - in stable/10/sys: amd64/amd64 arm/arm i386/i386 vm X-SVN-Group: stable-10 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-BeenThere: svn-src-stable@freebsd.org X-Mailman-Version: 2.1.18 Precedence: list List-Id: SVN commit messages for all the -stable branches of the src tree List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 24 Jul 2014 16:29:46 -0000 Author: kib Date: Thu Jul 24 16:29:44 2014 New Revision: 269072 URL: http://svnweb.freebsd.org/changeset/base/269072 Log: MFC r267213 (by alc): Add a page size field to struct vm_page. Approved by: alc Modified: stable/10/sys/amd64/amd64/pmap.c stable/10/sys/arm/arm/pmap-v6.c stable/10/sys/i386/i386/pmap.c stable/10/sys/vm/vm_map.c stable/10/sys/vm/vm_page.c stable/10/sys/vm/vm_page.h stable/10/sys/vm/vm_reserv.c Directory Properties: stable/10/ (props changed) Modified: stable/10/sys/amd64/amd64/pmap.c ============================================================================== --- stable/10/sys/amd64/amd64/pmap.c Thu Jul 24 15:49:28 2014 (r269071) +++ stable/10/sys/amd64/amd64/pmap.c Thu Jul 24 16:29:44 2014 (r269072) @@ -4437,9 +4437,7 @@ pmap_enter_object(pmap_t pmap, vm_offset while (m != NULL && (diff = m->pindex - m_start->pindex) < psize) { va = start + ptoa(diff); if ((va & PDRMASK) == 0 && va + NBPDR <= end && - (VM_PAGE_TO_PHYS(m) & PDRMASK) == 0 && - pmap_ps_enabled(pmap) && - vm_reserv_level_iffullpop(m) == 0 && + m->psind == 1 && pmap_ps_enabled(pmap) && pmap_enter_pde(pmap, va, m, prot, &lock)) m = &m[NBPDR / PAGE_SIZE - 1]; else Modified: stable/10/sys/arm/arm/pmap-v6.c ============================================================================== --- stable/10/sys/arm/arm/pmap-v6.c Thu Jul 24 15:49:28 2014 (r269071) +++ stable/10/sys/arm/arm/pmap-v6.c Thu Jul 24 16:29:44 2014 (r269072) @@ -3226,8 +3226,7 @@ pmap_enter_object(pmap_t pmap, vm_offset while (m != NULL && (diff = m->pindex - m_start->pindex) < psize) { va = start + ptoa(diff); if ((va & L1_S_OFFSET) == 0 && L2_NEXT_BUCKET(va) <= end && - (VM_PAGE_TO_PHYS(m) & L1_S_OFFSET) == 0 && - sp_enabled && vm_reserv_level_iffullpop(m) == 0 && + m->psind == 1 && sp_enabled && pmap_enter_section(pmap, va, m, prot)) m = &m[L1_S_SIZE / PAGE_SIZE - 1]; else Modified: stable/10/sys/i386/i386/pmap.c ============================================================================== --- stable/10/sys/i386/i386/pmap.c Thu Jul 24 15:49:28 2014 (r269071) +++ stable/10/sys/i386/i386/pmap.c Thu Jul 24 16:29:44 2014 (r269072) @@ -3733,8 +3733,7 @@ pmap_enter_object(pmap_t pmap, vm_offset while (m != NULL && (diff = m->pindex - m_start->pindex) < psize) { va = start + ptoa(diff); if ((va & PDRMASK) == 0 && va + NBPDR <= end && - (VM_PAGE_TO_PHYS(m) & PDRMASK) == 0 && - pg_ps_enabled && vm_reserv_level_iffullpop(m) == 0 && + m->psind == 1 && pg_ps_enabled && pmap_enter_pde(pmap, va, m, prot)) m = &m[NBPDR / PAGE_SIZE - 1]; else Modified: stable/10/sys/vm/vm_map.c ============================================================================== --- stable/10/sys/vm/vm_map.c Thu Jul 24 15:49:28 2014 (r269071) +++ stable/10/sys/vm/vm_map.c Thu Jul 24 16:29:44 2014 (r269072) @@ -1802,20 +1802,22 @@ vm_map_submap( } /* - * The maximum number of pages to map + * The maximum number of pages to map if MAP_PREFAULT_PARTIAL is specified */ #define MAX_INIT_PT 96 /* * vm_map_pmap_enter: * - * Preload read-only mappings for the specified object's resident pages - * into the target map. If "flags" is MAP_PREFAULT_PARTIAL, then only - * the resident pages within the address range [addr, addr + ulmin(size, - * ptoa(MAX_INIT_PT))) are mapped. Otherwise, all resident pages within - * the specified address range are mapped. This eliminates many soft - * faults on process startup and immediately after an mmap(2). Because - * these are speculative mappings, cached pages are not reactivated and + * Preload the specified map's pmap with mappings to the specified + * object's memory-resident pages. No further physical pages are + * allocated, and no further virtual pages are retrieved from secondary + * storage. If the specified flags include MAP_PREFAULT_PARTIAL, then a + * limited number of page mappings are created at the low-end of the + * specified address range. (For this purpose, a superpage mapping + * counts as one page mapping.) Otherwise, all resident pages within + * the specified address range are mapped. Because these mappings are + * being created speculatively, cached pages are not reactivated and * mapped. */ void @@ -1824,7 +1826,7 @@ vm_map_pmap_enter(vm_map_t map, vm_offse { vm_offset_t start; vm_page_t p, p_start; - vm_pindex_t psize, tmpidx; + vm_pindex_t mask, psize, threshold, tmpidx; if ((prot & (VM_PROT_READ | VM_PROT_EXECUTE)) == 0 || object == NULL) return; @@ -1842,8 +1844,6 @@ vm_map_pmap_enter(vm_map_t map, vm_offse } psize = atop(size); - if (psize > MAX_INIT_PT && (flags & MAP_PREFAULT_PARTIAL) != 0) - psize = MAX_INIT_PT; if (psize + pindex > object->size) { if (object->size < pindex) { VM_OBJECT_RUNLOCK(object); @@ -1854,6 +1854,7 @@ vm_map_pmap_enter(vm_map_t map, vm_offse start = 0; p_start = NULL; + threshold = MAX_INIT_PT; p = vm_page_find_least(object, pindex); /* @@ -1868,8 +1869,10 @@ vm_map_pmap_enter(vm_map_t map, vm_offse * don't allow an madvise to blow away our really * free pages allocating pv entries. */ - if ((flags & MAP_PREFAULT_MADVISE) && - cnt.v_free_count < cnt.v_free_reserved) { + if (((flags & MAP_PREFAULT_MADVISE) != 0 && + cnt.v_free_count < cnt.v_free_reserved) || + ((flags & MAP_PREFAULT_PARTIAL) != 0 && + tmpidx >= threshold)) { psize = tmpidx; break; } @@ -1878,6 +1881,16 @@ vm_map_pmap_enter(vm_map_t map, vm_offse start = addr + ptoa(tmpidx); p_start = p; } + /* Jump ahead if a superpage mapping is possible. */ + if (p->psind > 0 && ((addr + ptoa(tmpidx)) & + (pagesizes[p->psind] - 1)) == 0) { + mask = atop(pagesizes[p->psind]) - 1; + if (tmpidx + mask < psize && + vm_page_ps_is_valid(p)) { + p += mask; + threshold += mask; + } + } } else if (p_start != NULL) { pmap_enter_object(map->pmap, start, addr + ptoa(tmpidx), p_start, prot); Modified: stable/10/sys/vm/vm_page.c ============================================================================== --- stable/10/sys/vm/vm_page.c Thu Jul 24 15:49:28 2014 (r269071) +++ stable/10/sys/vm/vm_page.c Thu Jul 24 16:29:44 2014 (r269072) @@ -3058,6 +3058,31 @@ vm_page_is_valid(vm_page_t m, int base, } /* + * vm_page_ps_is_valid: + * + * Returns TRUE if the entire (super)page is valid and FALSE otherwise. + */ +boolean_t +vm_page_ps_is_valid(vm_page_t m) +{ + int i, npages; + + VM_OBJECT_ASSERT_LOCKED(m->object); + npages = atop(pagesizes[m->psind]); + + /* + * The physically contiguous pages that make up a superpage, i.e., a + * page with a page size index ("psind") greater than zero, will + * occupy adjacent entries in vm_page_array[]. + */ + for (i = 0; i < npages; i++) { + if (m[i].valid != VM_PAGE_BITS_ALL) + return (FALSE); + } + return (TRUE); +} + +/* * Set the page's dirty bits if the page is modified. */ void Modified: stable/10/sys/vm/vm_page.h ============================================================================== --- stable/10/sys/vm/vm_page.h Thu Jul 24 15:49:28 2014 (r269071) +++ stable/10/sys/vm/vm_page.h Thu Jul 24 16:29:44 2014 (r269072) @@ -157,6 +157,7 @@ struct vm_page { /* so, on normal X86 kernels, they must be at least 8 bits wide */ vm_page_bits_t valid; /* map of valid DEV_BSIZE chunks (O) */ vm_page_bits_t dirty; /* map of dirty DEV_BSIZE chunks (M) */ + int8_t psind; /* pagesizes[] index (O) */ }; /* @@ -450,6 +451,7 @@ vm_page_t vm_page_next(vm_page_t m); int vm_page_pa_tryrelock(pmap_t, vm_paddr_t, vm_paddr_t *); struct vm_pagequeue *vm_page_pagequeue(vm_page_t m); vm_page_t vm_page_prev(vm_page_t m); +boolean_t vm_page_ps_is_valid(vm_page_t m); void vm_page_putfake(vm_page_t m); void vm_page_readahead_finish(vm_page_t m); void vm_page_reference(vm_page_t m); Modified: stable/10/sys/vm/vm_reserv.c ============================================================================== --- stable/10/sys/vm/vm_reserv.c Thu Jul 24 15:49:28 2014 (r269071) +++ stable/10/sys/vm/vm_reserv.c Thu Jul 24 16:29:44 2014 (r269072) @@ -229,6 +229,11 @@ vm_reserv_depopulate(vm_reserv_t rv) if (rv->inpartpopq) { TAILQ_REMOVE(&vm_rvq_partpop, rv, partpopq); rv->inpartpopq = FALSE; + } else { + KASSERT(rv->pages->psind == 1, + ("vm_reserv_depopulate: reserv %p is already demoted", + rv)); + rv->pages->psind = 0; } rv->popcnt--; if (rv->popcnt == 0) { @@ -278,6 +283,8 @@ vm_reserv_populate(vm_reserv_t rv) ("vm_reserv_populate: reserv %p is free", rv)); KASSERT(rv->popcnt < VM_LEVEL_0_NPAGES, ("vm_reserv_populate: reserv %p is already full", rv)); + KASSERT(rv->pages->psind == 0, + ("vm_reserv_populate: reserv %p is already promoted", rv)); if (rv->inpartpopq) { TAILQ_REMOVE(&vm_rvq_partpop, rv, partpopq); rv->inpartpopq = FALSE; @@ -286,7 +293,8 @@ vm_reserv_populate(vm_reserv_t rv) if (rv->popcnt < VM_LEVEL_0_NPAGES) { rv->inpartpopq = TRUE; TAILQ_INSERT_TAIL(&vm_rvq_partpop, rv, partpopq); - } + } else + rv->pages->psind = 1; } /*