/* * (MPSAFE) * * Copyright (c) 1998-2010 The DragonFly Project. All rights reserved. * * This code is derived from software contributed to The DragonFly Project * by Matthew Dillon <dillon@backplane.com> * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the * distribution. * 3. Neither the name of The DragonFly Project nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific, prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * Copyright (c) 1994 John S. Dyson * Copyright (c) 1990 University of Utah. * Copyright (c) 1991, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * the Systems Programming Group of the University of Utah Computer * Science Department. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * New Swap System * Matthew Dillon * * Radix Bitmap 'blists'. * * - The new swapper uses the new radix bitmap code. This should scale * to arbitrarily small or arbitrarily large swap spaces and an almost * arbitrary degree of fragmentation. * * Features: * * - on the fly reallocation of swap during putpages. The new system * does not try to keep previously allocated swap blocks for dirty * pages. * * - on the fly deallocation of swap * * - No more garbage collection required. Unnecessarily allocated swap * blocks only exist for dirty vm_page_t's now and these are already * cycled (in a high-load system) by the pager. We also do on-the-fly * removal of invalidated swap blocks when a page is destroyed * or renamed. * * from: Utah $Hdr: swap_pager.c 1.4 91/04/30$ * @(#)swap_pager.c 8.9 (Berkeley) 3/21/94 * $FreeBSD: src/sys/vm/swap_pager.c,v 1.130.2.12 2002/08/31 21:15:55 dillon Exp $ */ #include "opt_swap.h" #include <sys/param.h> #include <sys/systm.h> #include <sys/conf.h> #include <sys/kernel.h> #include <sys/proc.h> #include <sys/buf.h> #include <sys/vnode.h> #include <sys/malloc.h> #include <sys/vmmeter.h> #include <sys/sysctl.h> #include <sys/blist.h> #include <sys/lock.h> #include <sys/kcollect.h> #include <vm/vm.h> #include <vm/vm_object.h> #include <vm/vm_page.h> #include <vm/vm_pager.h> #include <vm/vm_pageout.h> #include <vm/swap_pager.h> #include <vm/vm_extern.h> #include <vm/vm_zone.h> #include <vm/vnode_pager.h> #include <sys/buf2.h> #include <vm/vm_page2.h> #ifndef MAX_PAGEOUT_CLUSTER #define MAX_PAGEOUT_CLUSTER SWB_NPAGES #endif #define SWM_FREE 0x02 /* free, period */ #define SWM_POP 0x04 /* pop out */ #define SWBIO_READ 0x01 #define SWBIO_WRITE 0x02 #define SWBIO_SYNC 0x04 #define SWBIO_TTC 0x08 /* for OBJPC_TRY_TO_CACHE */ struct swfreeinfo { vm_object_t object; vm_pindex_t basei; vm_pindex_t begi; vm_pindex_t endi; /* inclusive */ }; struct swswapoffinfo { vm_object_t object; int devidx; int shared; }; /* * vm_swap_size is in page-sized chunks now. It was DEV_BSIZE'd chunks * in the old system. */ int swap_pager_full; /* swap space exhaustion (task killing) */ int swap_fail_ticks; /* when we became exhausted */ int swap_pager_almost_full; /* swap space exhaustion (w/ hysteresis)*/ swblk_t vm_swap_cache_use; swblk_t vm_swap_anon_use; static int vm_report_swap_allocs; static struct krate kswaprate = { 1 }; static int nsw_rcount; /* free read buffers */ static int nsw_wcount_sync; /* limit write buffers / synchronous */ static int nsw_wcount_async; /* limit write buffers / asynchronous */ static int nsw_wcount_async_max;/* assigned maximum */ static int nsw_cluster_max; /* maximum VOP I/O allowed */ struct blist *swapblist; static int swap_async_max = 4; /* maximum in-progress async I/O's */ static int swap_burst_read = 0; /* allow burst reading */ static swblk_t swapiterator; /* linearize allocations */ int swap_user_async = 0; /* user swap pager operation can be async */ static struct spinlock swapbp_spin = SPINLOCK_INITIALIZER(&swapbp_spin, "swapbp_spin"); /* from vm_swap.c */ extern struct vnode *swapdev_vp; extern struct swdevt *swdevt; extern int nswdev; #define BLK2DEVIDX(blk) (nswdev > 1 ? blk / SWB_DMMAX % nswdev : 0) SYSCTL_INT(_vm, OID_AUTO, swap_async_max, CTLFLAG_RW, &swap_async_max, 0, "Maximum running async swap ops"); SYSCTL_INT(_vm, OID_AUTO, swap_burst_read, CTLFLAG_RW, &swap_burst_read, 0, "Allow burst reads for pageins"); SYSCTL_INT(_vm, OID_AUTO, swap_user_async, CTLFLAG_RW, &swap_user_async, 0, "Allow async uuser swap write I/O"); #if SWBLK_BITS == 64 SYSCTL_LONG(_vm, OID_AUTO, swap_cache_use, CTLFLAG_RD, &vm_swap_cache_use, 0, ""); SYSCTL_LONG(_vm, OID_AUTO, swap_anon_use, CTLFLAG_RD, &vm_swap_anon_use, 0, ""); SYSCTL_LONG(_vm, OID_AUTO, swap_free, CTLFLAG_RD, &vm_swap_size, 0, ""); SYSCTL_LONG(_vm, OID_AUTO, swap_size, CTLFLAG_RD, &vm_swap_max, 0, ""); #else SYSCTL_INT(_vm, OID_AUTO, swap_cache_use, CTLFLAG_RD, &vm_swap_cache_use, 0, ""); SYSCTL_INT(_vm, OID_AUTO, swap_anon_use, CTLFLAG_RD, &vm_swap_anon_use, 0, ""); SYSCTL_INT(_vm, OID_AUTO, swap_free, CTLFLAG_RD, &vm_swap_size, 0, ""); SYSCTL_INT(_vm, OID_AUTO, swap_size, CTLFLAG_RD, &vm_swap_max, 0, ""); #endif SYSCTL_INT(_vm, OID_AUTO, report_swap_allocs, CTLFLAG_RW, &vm_report_swap_allocs, 0, ""); __read_mostly vm_zone_t swap_zone; /* * Red-Black tree for swblock entries * * The caller must hold vm_token */ RB_GENERATE2(swblock_rb_tree, swblock, swb_entry, rb_swblock_compare, vm_pindex_t, swb_index); int rb_swblock_compare(struct swblock *swb1, struct swblock *swb2) { if (swb1->swb_index < swb2->swb_index) return(-1); if (swb1->swb_index > swb2->swb_index) return(1); return(0); } static int rb_swblock_scancmp(struct swblock *swb, void *data) { struct swfreeinfo *info = data; if (swb->swb_index < info->basei) return(-1); if (swb->swb_index > info->endi) return(1); return(0); } static int rb_swblock_condcmp(struct swblock *swb, void *data) { struct swfreeinfo *info = data; if (swb->swb_index < info->basei) return(-1); return(0); } /* * pagerops for OBJT_SWAP - "swap pager". Some ops are also global procedure * calls hooked from other parts of the VM system and do not appear here. * (see vm/swap_pager.h). */ static void swap_pager_dealloc (vm_object_t object); static int swap_pager_getpage (vm_object_t, vm_pindex_t, vm_page_t *, int); static void swap_chain_iodone(struct bio *biox); struct pagerops swappagerops = { swap_pager_dealloc, /* deallocate an OBJT_SWAP object */ swap_pager_getpage, /* pagein */ swap_pager_putpages, /* pageout */ swap_pager_haspage /* get backing store status for page */ }; /* * SWB_DMMAX is in page-sized chunks with the new swap system. It was * dev-bsized chunks in the old. SWB_DMMAX is always a power of 2. * * swap_*() routines are externally accessible. swp_*() routines are * internal. */ int nswap_lowat = 128; /* in pages, swap_pager_almost_full warn */ int nswap_hiwat = 512; /* in pages, swap_pager_almost_full warn */ static __inline void swp_sizecheck (void); static void swp_pager_async_iodone (struct bio *bio); /* * Swap bitmap functions */ static __inline void swp_pager_freeswapspace(vm_object_t object, swblk_t blk, int npages); static __inline swblk_t swp_pager_getswapspace(vm_object_t object, int npages); /* * Metadata functions */ static void swp_pager_meta_convert(vm_object_t); static void swp_pager_meta_build(vm_object_t, vm_pindex_t, swblk_t); static void swp_pager_meta_free(vm_object_t, vm_pindex_t, vm_pindex_t); static void swp_pager_meta_free_all(vm_object_t); static swblk_t swp_pager_meta_ctl(vm_object_t, vm_pindex_t, int); /* * SWP_SIZECHECK() - update swap_pager_full indication * * update the swap_pager_almost_full indication and warn when we are * about to run out of swap space, using lowat/hiwat hysteresis. * * Clear swap_pager_full ( task killing ) indication when lowat is met. * * No restrictions on call * This routine may not block. * SMP races are ok. */ static __inline void swp_sizecheck(void) { if (vm_swap_size < nswap_lowat) { if (swap_pager_almost_full == 0) { kprintf("swap_pager: out of swap space\n"); swap_pager_almost_full = 1; swap_fail_ticks = ticks; } } else { swap_pager_full = 0; if (vm_swap_size > nswap_hiwat) swap_pager_almost_full = 0; } } /* * Long-term data collection on 10-second interval. Return the value * for KCOLLECT_SWAPPCT and set the values for SWAPANO and SWAPCCAC. * * Return total swap in the scale field. This can change if swap is * regularly added or removed and may cause some historical confusion * in that case, but SWAPPCT will always be historically accurate. */ #define PTOB(value) ((uint64_t)(value) << PAGE_SHIFT) static uint64_t collect_swap_callback(int n) { uint64_t total = vm_swap_max; uint64_t anon = vm_swap_anon_use; uint64_t cache = vm_swap_cache_use; if (total == 0) /* avoid divide by zero */ total = 1; kcollect_setvalue(KCOLLECT_SWAPANO, PTOB(anon)); kcollect_setvalue(KCOLLECT_SWAPCAC, PTOB(cache)); kcollect_setscale(KCOLLECT_SWAPANO, KCOLLECT_SCALE(KCOLLECT_SWAPANO_FORMAT, PTOB(total))); kcollect_setscale(KCOLLECT_SWAPCAC, KCOLLECT_SCALE(KCOLLECT_SWAPCAC_FORMAT, PTOB(total))); return (((anon + cache) * 10000 + (total >> 1)) / total); } /* * SWAP_PAGER_INIT() - initialize the swap pager! * * Expected to be started from system init. NOTE: This code is run * before much else so be careful what you depend on. Most of the VM * system has yet to be initialized at this point. * * Called from the low level boot code only. */ static void swap_pager_init(void *arg __unused) { kcollect_register(KCOLLECT_SWAPPCT, "swapuse", collect_swap_callback, KCOLLECT_SCALE(KCOLLECT_SWAPPCT_FORMAT, 0)); kcollect_register(KCOLLECT_SWAPANO, "swapano", NULL, KCOLLECT_SCALE(KCOLLECT_SWAPANO_FORMAT, 0)); kcollect_register(KCOLLECT_SWAPCAC, "swapcac", NULL, KCOLLECT_SCALE(KCOLLECT_SWAPCAC_FORMAT, 0)); } SYSINIT(vm_mem, SI_BOOT1_VM, SI_ORDER_THIRD, swap_pager_init, NULL); /* * SWAP_PAGER_SWAP_INIT() - swap pager initialization from pageout process * * Expected to be started from pageout process once, prior to entering * its main loop. * * Called from the low level boot code only. */ void swap_pager_swap_init(void) { int n, n2; /* * Number of in-transit swap bp operations. Don't * exhaust the pbufs completely. Make sure we * initialize workable values (0 will work for hysteresis * but it isn't very efficient). * * The nsw_cluster_max is constrained by the number of pages an XIO * holds, i.e., (MAXPHYS/PAGE_SIZE) and our locally defined * MAX_PAGEOUT_CLUSTER. Also be aware that swap ops are * constrained by the swap device interleave stripe size. * * Currently we hardwire nsw_wcount_async to 4. This limit is * designed to prevent other I/O from having high latencies due to * our pageout I/O. The value 4 works well for one or two active swap * devices but is probably a little low if you have more. Even so, * a higher value would probably generate only a limited improvement * with three or four active swap devices since the system does not * typically have to pageout at extreme bandwidths. We will want * at least 2 per swap devices, and 4 is a pretty good value if you * have one NFS swap device due to the command/ack latency over NFS. * So it all works out pretty well. */ nsw_cluster_max = min((MAXPHYS/PAGE_SIZE), MAX_PAGEOUT_CLUSTER); nsw_rcount = (nswbuf_kva + 1) / 2; nsw_wcount_sync = (nswbuf_kva + 3) / 4; nsw_wcount_async = 4; nsw_wcount_async_max = nsw_wcount_async; /* * The zone is dynamically allocated so generally size it to * maxswzone (32MB to 256GB of KVM). Set a minimum size based * on physical memory of around 8x (each swblock can hold 16 pages). * * With the advent of SSDs (vs HDs) the practical (swap:memory) ratio * has increased dramatically. */ n = vmstats.v_page_count / 2; if (maxswzone && n < maxswzone / sizeof(struct swblock)) n = maxswzone / sizeof(struct swblock); n2 = n; do { swap_zone = zinit( "SWAPMETA", sizeof(struct swblock), n, ZONE_INTERRUPT); if (swap_zone != NULL) break; /* * if the allocation failed, try a zone two thirds the * size of the previous attempt. */ n -= ((n + 2) / 3); } while (n > 0); if (swap_zone == NULL) panic("swap_pager_swap_init: swap_zone == NULL"); if (n2 != n) kprintf("Swap zone entries reduced from %d to %d.\n", n2, n); } /* * SWAP_PAGER_ALLOC() - allocate a new OBJT_SWAP VM object and instantiate * its metadata structures. * * This routine is called from the mmap and fork code to create a new * OBJT_SWAP object. We do this by creating an OBJT_DEFAULT object * and then converting it with swp_pager_meta_convert(). * * We only support unnamed objects. * * No restrictions. */ vm_object_t swap_pager_alloc(void *handle, off_t size, vm_prot_t prot, off_t offset) { vm_object_t object; KKASSERT(handle == NULL); object = vm_object_allocate_hold(OBJT_DEFAULT, OFF_TO_IDX(offset + PAGE_MASK + size)); swp_pager_meta_convert(object); vm_object_drop(object); return (object); } /* * SWAP_PAGER_DEALLOC() - remove swap metadata from object * * The swap backing for the object is destroyed. The code is * designed such that we can reinstantiate it later, but this * routine is typically called only when the entire object is * about to be destroyed. * * The object must be locked or unreferenceable. * No other requirements. */ static void swap_pager_dealloc(vm_object_t object) { vm_object_hold(object); vm_object_pip_wait(object, "swpdea"); /* * Free all remaining metadata. We only bother to free it from * the swap meta data. We do not attempt to free swapblk's still * associated with vm_page_t's for this object. We do not care * if paging is still in progress on some objects. */ swp_pager_meta_free_all(object); vm_object_drop(object); } /************************************************************************ * SWAP PAGER BITMAP ROUTINES * ************************************************************************/ /* * SWP_PAGER_GETSWAPSPACE() - allocate raw swap space * * Allocate swap for the requested number of pages. The starting * swap block number (a page index) is returned or SWAPBLK_NONE * if the allocation failed. * * Also has the side effect of advising that somebody made a mistake * when they configured swap and didn't configure enough. * * The caller must hold the object. * This routine may not block. */ static __inline swblk_t swp_pager_getswapspace(vm_object_t object, int npages) { swblk_t blk; lwkt_gettoken(&vm_token); blk = blist_allocat(swapblist, npages, swapiterator); if (blk == SWAPBLK_NONE) blk = blist_allocat(swapblist, npages, 0); if (blk == SWAPBLK_NONE) { if (swap_pager_full != 2) { if (vm_swap_max == 0) { krateprintf(&kswaprate, "Warning: The system would like to " "page to swap but no swap space " "is configured!\n"); } else { krateprintf(&kswaprate, "swap_pager_getswapspace: " "swap full allocating %d pages\n", npages); } swap_pager_full = 2; if (swap_pager_almost_full == 0) swap_fail_ticks = ticks; swap_pager_almost_full = 1; } } else { /* swapiterator = blk; disable for now, doesn't work well */ swapacctspace(blk, -npages); if (object->type == OBJT_SWAP) vm_swap_anon_use += npages; else vm_swap_cache_use += npages; swp_sizecheck(); } lwkt_reltoken(&vm_token); return(blk); } /* * SWP_PAGER_FREESWAPSPACE() - free raw swap space * * This routine returns the specified swap blocks back to the bitmap. * * Note: This routine may not block (it could in the old swap code), * and through the use of the new blist routines it does not block. * * This routine may not block. */ static __inline void swp_pager_freeswapspace(vm_object_t object, swblk_t blk, int npages) { struct swdevt *sp = &swdevt[BLK2DEVIDX(blk)]; lwkt_gettoken(&vm_token); sp->sw_nused -= npages; if (object->type == OBJT_SWAP) vm_swap_anon_use -= npages; else vm_swap_cache_use -= npages; if (sp->sw_flags & SW_CLOSING) { lwkt_reltoken(&vm_token); return; } blist_free(swapblist, blk, npages); vm_swap_size += npages; swp_sizecheck(); lwkt_reltoken(&vm_token); } /* * SWAP_PAGER_FREESPACE() - frees swap blocks associated with a page * range within an object. * * This is a globally accessible routine. * * This routine removes swapblk assignments from swap metadata. * * The external callers of this routine typically have already destroyed * or renamed vm_page_t's associated with this range in the object so * we should be ok. * * No requirements. */ void swap_pager_freespace(vm_object_t object, vm_pindex_t start, vm_pindex_t size) { if (object->swblock_count == 0) return; vm_object_hold(object); swp_pager_meta_free(object, start, size); vm_object_drop(object); } /* * No requirements. */ void swap_pager_freespace_all(vm_object_t object) { if (object->swblock_count == 0) return; vm_object_hold(object); swp_pager_meta_free_all(object); vm_object_drop(object); } /* * This function conditionally frees swap cache swap starting at * (*basei) in the object. (count) swap blocks will be nominally freed. * The actual number of blocks freed can be more or less than the * requested number. * * This function nominally returns the number of blocks freed. However, * the actual number of blocks freed may be less then the returned value. * If the function is unable to exhaust the object or if it is able to * free (approximately) the requested number of blocks it returns * a value n > count. * * If we exhaust the object we will return a value n <= count. * * The caller must hold the object. * * WARNING! If count == 0 then -1 can be returned as a degenerate case, * callers should always pass a count value > 0. */ static int swap_pager_condfree_callback(struct swblock *swap, void *data); int swap_pager_condfree(vm_object_t object, vm_pindex_t *basei, int count) { struct swfreeinfo info; int n; int t; ASSERT_LWKT_TOKEN_HELD(vm_object_token(object)); info.object = object; info.basei = *basei; /* skip up to this page index */ info.begi = count; /* max swap pages to destroy */ info.endi = count * 8; /* max swblocks to scan */ swblock_rb_tree_RB_SCAN(&object->swblock_root, rb_swblock_condcmp, swap_pager_condfree_callback, &info); *basei = info.basei; /* * Take the higher difference swblocks vs pages */ n = count - (int)info.begi; t = count * 8 - (int)info.endi; if (n < t) n = t; if (n < 1) n = 1; return(n); } /* * The idea is to free whole meta-block to avoid fragmenting * the swap space or disk I/O. We only do this if NO VM pages * are present. * * We do not have to deal with clearing PG_SWAPPED in related VM * pages because there are no related VM pages. * * The caller must hold the object. */ static int swap_pager_condfree_callback(struct swblock *swap, void *data) { struct swfreeinfo *info = data; vm_object_t object = info->object; int i; for (i = 0; i < SWAP_META_PAGES; ++i) { if (vm_page_lookup(object, swap->swb_index + i)) break; } info->basei = swap->swb_index + SWAP_META_PAGES; if (i == SWAP_META_PAGES) { info->begi -= swap->swb_count; swap_pager_freespace(object, swap->swb_index, SWAP_META_PAGES); } --info->endi; if ((int)info->begi < 0 || (int)info->endi < 0) return(-1); lwkt_yield(); return(0); } /* * Called by vm_page_alloc() when a new VM page is inserted * into a VM object. Checks whether swap has been assigned to * the page and sets PG_SWAPPED as necessary. * * (m) must be busied by caller and remains busied on return. */ void swap_pager_page_inserted(vm_page_t m) { if (m->object->swblock_count) { vm_object_hold(m->object); if (swp_pager_meta_ctl(m->object, m->pindex, 0) != SWAPBLK_NONE) vm_page_flag_set(m, PG_SWAPPED); vm_object_drop(m->object); } } /* * SWAP_PAGER_RESERVE() - reserve swap blocks in object * * Assigns swap blocks to the specified range within the object. The * swap blocks are not zerod. Any previous swap assignment is destroyed. * * Returns 0 on success, -1 on failure. * * The caller is responsible for avoiding races in the specified range. * No other requirements. */ int swap_pager_reserve(vm_object_t object, vm_pindex_t start, vm_size_t size) { int n = 0; swblk_t blk = SWAPBLK_NONE; vm_pindex_t beg = start; /* save start index */ vm_object_hold(object); while (size) { if (n == 0) { n = BLIST_MAX_ALLOC; while ((blk = swp_pager_getswapspace(object, n)) == SWAPBLK_NONE) { n >>= 1; if (n == 0) { swp_pager_meta_free(object, beg, start - beg); vm_object_drop(object); return(-1); } } } swp_pager_meta_build(object, start, blk); --size; ++start; ++blk; --n; } swp_pager_meta_free(object, start, n); vm_object_drop(object); return(0); } /* * SWAP_PAGER_COPY() - copy blocks from source pager to destination pager * and destroy the source. * * Copy any valid swapblks from the source to the destination. In * cases where both the source and destination have a valid swapblk, * we keep the destination's. * * This routine is allowed to block. It may block allocating metadata * indirectly through swp_pager_meta_build() or if paging is still in * progress on the source. * * XXX vm_page_collapse() kinda expects us not to block because we * supposedly do not need to allocate memory, but for the moment we * *may* have to get a little memory from the zone allocator, but * it is taken from the interrupt memory. We should be ok. * * The source object contains no vm_page_t's (which is just as well) * The source object is of type OBJT_SWAP. * * The source and destination objects must be held by the caller. */ void swap_pager_copy(vm_object_t srcobject, vm_object_t dstobject, vm_pindex_t base_index, int destroysource) { vm_pindex_t i; ASSERT_LWKT_TOKEN_HELD(vm_object_token(srcobject)); ASSERT_LWKT_TOKEN_HELD(vm_object_token(dstobject)); /* * transfer source to destination. */ for (i = 0; i < dstobject->size; ++i) { swblk_t dstaddr; /* * Locate (without changing) the swapblk on the destination, * unless it is invalid in which case free it silently, or * if the destination is a resident page, in which case the * source is thrown away. */ dstaddr = swp_pager_meta_ctl(dstobject, i, 0); if (dstaddr == SWAPBLK_NONE) { /* * Destination has no swapblk and is not resident, * copy source. */ swblk_t srcaddr; srcaddr = swp_pager_meta_ctl(srcobject, base_index + i, SWM_POP); if (srcaddr != SWAPBLK_NONE) swp_pager_meta_build(dstobject, i, srcaddr); } else { /* * Destination has valid swapblk or it is represented * by a resident page. We destroy the sourceblock. */ swp_pager_meta_ctl(srcobject, base_index + i, SWM_FREE); } } /* * Free left over swap blocks in source. * * We have to revert the type to OBJT_DEFAULT so we do not accidently * double-remove the object from the swap queues. */ if (destroysource) { /* * Reverting the type is not necessary, the caller is going * to destroy srcobject directly, but I'm doing it here * for consistency since we've removed the object from its * queues. */ swp_pager_meta_free_all(srcobject); if (srcobject->type == OBJT_SWAP) srcobject->type = OBJT_DEFAULT; } } /* * SWAP_PAGER_HASPAGE() - determine if we have good backing store for * the requested page. * * We determine whether good backing store exists for the requested * page and return TRUE if it does, FALSE if it doesn't. * * If TRUE, we also try to determine how much valid, contiguous backing * store exists before and after the requested page within a reasonable * distance. We do not try to restrict it to the swap device stripe * (that is handled in getpages/putpages). It probably isn't worth * doing here. * * No requirements. */ boolean_t swap_pager_haspage(vm_object_t object, vm_pindex_t pindex) { swblk_t blk0; /* * do we have good backing store at the requested index ? */ vm_object_hold(object); blk0 = swp_pager_meta_ctl(object, pindex, 0); if (blk0 == SWAPBLK_NONE) { vm_object_drop(object); return (FALSE); } vm_object_drop(object); return (TRUE); } /* * Object must be held exclusive or shared by the caller. */ boolean_t swap_pager_haspage_locked(vm_object_t object, vm_pindex_t pindex) { if (swp_pager_meta_ctl(object, pindex, 0) == SWAPBLK_NONE) return FALSE; return TRUE; } /* * SWAP_PAGER_PAGE_UNSWAPPED() - remove swap backing store related to page * * This removes any associated swap backing store, whether valid or * not, from the page. This operates on any VM object, not just OBJT_SWAP * objects. * * This routine is typically called when a page is made dirty, at * which point any associated swap can be freed. MADV_FREE also * calls us in a special-case situation * * NOTE!!! If the page is clean and the swap was valid, the caller * should make the page dirty before calling this routine. * This routine does NOT change the m->dirty status of the page. * Also: MADV_FREE depends on it. * * The page must be busied. * The caller can hold the object to avoid blocking, else we might block. * No other requirements. */ void swap_pager_unswapped(vm_page_t m) { if (m->flags & PG_SWAPPED) { vm_object_hold(m->object); KKASSERT(m->flags & PG_SWAPPED); swp_pager_meta_ctl(m->object, m->pindex, SWM_FREE); vm_page_flag_clear(m, PG_SWAPPED); vm_object_drop(m->object); } } /* * SWAP_PAGER_STRATEGY() - read, write, free blocks * * This implements a VM OBJECT strategy function using swap backing store. * This can operate on any VM OBJECT type, not necessarily just OBJT_SWAP * types. Only BUF_CMD_{READ,WRITE,FREEBLKS} is supported, any other * requests will return EINVAL. * * This is intended to be a cacheless interface (i.e. caching occurs at * higher levels), and is also used as a swap-based SSD cache for vnode * and device objects. * * All I/O goes directly to and from the swap device. * * We currently attempt to run I/O synchronously or asynchronously as * the caller requests. This isn't perfect because we loose error * sequencing when we run multiple ops in parallel to satisfy a request. * But this is swap, so we let it all hang out. * * NOTE: This function supports the KVABIO API wherein bp->b_data might * not be synchronized to the current cpu. * * No requirements. */ void swap_pager_strategy(vm_object_t object, struct bio *bio) { struct buf *bp = bio->bio_buf; struct bio *nbio; vm_pindex_t start; vm_pindex_t biox_blkno = 0; int count; char *data; struct bio *biox; struct buf *bufx; #if 0 struct bio_track *track; #endif #if 0 /* * tracking for swapdev vnode I/Os */ if (bp->b_cmd == BUF_CMD_READ) track = &swapdev_vp->v_track_read; else track = &swapdev_vp->v_track_write; #endif /* * Only supported commands */ if (bp->b_cmd != BUF_CMD_FREEBLKS && bp->b_cmd != BUF_CMD_READ && bp->b_cmd != BUF_CMD_WRITE) { bp->b_error = EINVAL; bp->b_flags |= B_ERROR | B_INVAL; biodone(bio); return; } /* * bcount must be an integral number of pages. */ if (bp->b_bcount & PAGE_MASK) { bp->b_error = EINVAL; bp->b_flags |= B_ERROR | B_INVAL; biodone(bio); kprintf("swap_pager_strategy: bp %p offset %lld size %d, " "not page bounded\n", bp, (long long)bio->bio_offset, (int)bp->b_bcount); return; } /* * Clear error indication, initialize page index, count, data pointer. */ bp->b_error = 0; bp->b_flags &= ~B_ERROR; bp->b_resid = bp->b_bcount; start = (vm_pindex_t)(bio->bio_offset >> PAGE_SHIFT); count = howmany(bp->b_bcount, PAGE_SIZE); /* * WARNING! Do not dereference *data without issuing a bkvasync() */ data = bp->b_data; /* * Deal with BUF_CMD_FREEBLKS */ if (bp->b_cmd == BUF_CMD_FREEBLKS) { /* * FREE PAGE(s) - destroy underlying swap that is no longer * needed. */ vm_object_hold(object); swp_pager_meta_free(object, start, count); vm_object_drop(object); bp->b_resid = 0; biodone(bio); return; } /* * We need to be able to create a new cluster of I/O's. We cannot * use the caller fields of the passed bio so push a new one. * * Because nbio is just a placeholder for the cluster links, * we can biodone() the original bio instead of nbio to make * things a bit more efficient. */ nbio = push_bio(bio); nbio->bio_offset = bio->bio_offset; nbio->bio_caller_info1.cluster_head = NULL; nbio->bio_caller_info2.cluster_tail = NULL; biox = NULL; bufx = NULL; /* * Execute read or write */ vm_object_hold(object); while (count > 0) { swblk_t blk; /* * Obtain block. If block not found and writing, allocate a * new block and build it into the object. */ blk = swp_pager_meta_ctl(object, start, 0); if ((blk == SWAPBLK_NONE) && bp->b_cmd == BUF_CMD_WRITE) { blk = swp_pager_getswapspace(object, 1); if (blk == SWAPBLK_NONE) { bp->b_error = ENOMEM; bp->b_flags |= B_ERROR; break; } swp_pager_meta_build(object, start, blk); } /* * Do we have to flush our current collection? Yes if: * * - no swap block at this index * - swap block is not contiguous * - we cross a physical disk boundry in the * stripe. */ if (biox && (biox_blkno + btoc(bufx->b_bcount) != blk || ((biox_blkno ^ blk) & ~SWB_DMMASK))) { switch(bp->b_cmd) { case BUF_CMD_READ: ++mycpu->gd_cnt.v_swapin; mycpu->gd_cnt.v_swappgsin += btoc(bufx->b_bcount); break; case BUF_CMD_WRITE: ++mycpu->gd_cnt.v_swapout; mycpu->gd_cnt.v_swappgsout += btoc(bufx->b_bcount); bufx->b_dirtyend = bufx->b_bcount; break; default: /* NOT REACHED */ break; } /* * Finished with this buf. */ KKASSERT(bufx->b_bcount != 0); if (bufx->b_cmd != BUF_CMD_READ) bufx->b_dirtyend = bufx->b_bcount; biox = NULL; bufx = NULL; } /* * Add new swapblk to biox, instantiating biox if necessary. * Zero-fill reads are able to take a shortcut. */ if (blk == SWAPBLK_NONE) { /* * We can only get here if we are reading. */ bkvasync(bp); bzero(data, PAGE_SIZE); bp->b_resid -= PAGE_SIZE; } else { if (biox == NULL) { /* XXX chain count > 4, wait to <= 4 */ bufx = getpbuf(NULL); bufx->b_flags |= B_KVABIO; biox = &bufx->b_bio1; cluster_append(nbio, bufx); bufx->b_cmd = bp->b_cmd; biox->bio_done = swap_chain_iodone; biox->bio_offset = (off_t)blk << PAGE_SHIFT; biox->bio_caller_info1.cluster_parent = nbio; biox_blkno = blk; bufx->b_bcount = 0; bufx->b_data = data; } bufx->b_bcount += PAGE_SIZE; } --count; ++start; data += PAGE_SIZE; } vm_object_drop(object); /* * Flush out last buffer */ if (biox) { if (bufx->b_cmd == BUF_CMD_READ) { ++mycpu->gd_cnt.v_swapin; mycpu->gd_cnt.v_swappgsin += btoc(bufx->b_bcount); } else { ++mycpu->gd_cnt.v_swapout; mycpu->gd_cnt.v_swappgsout += btoc(bufx->b_bcount); bufx->b_dirtyend = bufx->b_bcount; } KKASSERT(bufx->b_bcount); if (bufx->b_cmd != BUF_CMD_READ) bufx->b_dirtyend = bufx->b_bcount; /* biox, bufx = NULL */ } /* * Now initiate all the I/O. Be careful looping on our chain as * I/O's may complete while we are still initiating them. * * If the request is a 100% sparse read no bios will be present * and we just biodone() the buffer. */ nbio->bio_caller_info2.cluster_tail = NULL; bufx = nbio->bio_caller_info1.cluster_head; if (bufx) { while (bufx) { biox = &bufx->b_bio1; BUF_KERNPROC(bufx); bufx = bufx->b_cluster_next; vn_strategy(swapdev_vp, biox); } } else { biodone(bio); } /* * Completion of the cluster will also call biodone_chain(nbio). * We never call biodone(nbio) so we don't have to worry about * setting up a bio_done callback. It's handled in the sub-IO. */ /**/ } /* * biodone callback * * No requirements. */ static void swap_chain_iodone(struct bio *biox) { struct buf **nextp; struct buf *bufx; /* chained sub-buffer */ struct bio *nbio; /* parent nbio with chain glue */ struct buf *bp; /* original bp associated with nbio */ int chain_empty; bufx = biox->bio_buf; nbio = biox->bio_caller_info1.cluster_parent; bp = nbio->bio_buf; /* * Update the original buffer */ KKASSERT(bp != NULL); if (bufx->b_flags & B_ERROR) { atomic_set_int(&bufx->b_flags, B_ERROR); bp->b_error = bufx->b_error; /* race ok */ } else if (bufx->b_resid != 0) { atomic_set_int(&bufx->b_flags, B_ERROR); bp->b_error = EINVAL; /* race ok */ } else { atomic_subtract_int(&bp->b_resid, bufx->b_bcount); } /* * Remove us from the chain. */ spin_lock(&swapbp_spin); nextp = &nbio->bio_caller_info1.cluster_head; while (*nextp != bufx) { KKASSERT(*nextp != NULL); nextp = &(*nextp)->b_cluster_next; } *nextp = bufx->b_cluster_next; chain_empty = (nbio->bio_caller_info1.cluster_head == NULL); spin_unlock(&swapbp_spin); /* * Clean up bufx. If the chain is now empty we finish out * the parent. Note that we may be racing other completions * so we must use the chain_empty status from above. */ if (chain_empty) { if (bp->b_resid != 0 && !(bp->b_flags & B_ERROR)) { atomic_set_int(&bp->b_flags, B_ERROR); bp->b_error = EINVAL; } biodone_chain(nbio); } relpbuf(bufx, NULL); } /* * SWAP_PAGER_GETPAGES() - bring page in from swap * * The requested page may have to be brought in from swap. Calculate the * swap block and bring in additional pages if possible. All pages must * have contiguous swap block assignments and reside in the same object. * * The caller has a single vm_object_pip_add() reference prior to * calling us and we should return with the same. * * The caller has BUSY'd the page. We should return with (*mpp) left busy, * and any additinal pages unbusied. * * If the caller encounters a PG_RAM page it will pass it to us even though * it may be valid and dirty. We cannot overwrite the page in this case! * The case is used to allow us to issue pure read-aheads. * * NOTE! XXX This code does not entirely pipeline yet due to the fact that * the PG_RAM page is validated at the same time as mreq. What we * really need to do is issue a separate read-ahead pbuf. * * No requirements. */ static int swap_pager_getpage(vm_object_t object, vm_pindex_t pindex, vm_page_t *mpp, int seqaccess) { struct buf *bp; struct bio *bio; vm_page_t mreq; vm_page_t m; vm_offset_t kva; swblk_t blk; int i; int j; int raonly; int error; u_int32_t busy_count; vm_page_t marray[XIO_INTERNAL_PAGES]; mreq = *mpp; vm_object_hold(object); if (mreq->object != object) { panic("swap_pager_getpages: object mismatch %p/%p", object, mreq->object ); } /* * We don't want to overwrite a fully valid page as it might be * dirty. This case can occur when e.g. vm_fault hits a perfectly * valid page with PG_RAM set. * * In this case we see if the next page is a suitable page-in * candidate and if it is we issue read-ahead. PG_RAM will be * set on the last page of the read-ahead to continue the pipeline. */ if (mreq->valid == VM_PAGE_BITS_ALL) { if (swap_burst_read == 0 || mreq->pindex + 1 >= object->size) { vm_object_drop(object); return(VM_PAGER_OK); } blk = swp_pager_meta_ctl(object, mreq->pindex + 1, 0); if (blk == SWAPBLK_NONE) { vm_object_drop(object); return(VM_PAGER_OK); } m = vm_page_lookup_busy_try(object, mreq->pindex + 1, TRUE, &error); if (error) { vm_object_drop(object); return(VM_PAGER_OK); } else if (m == NULL) { /* * Use VM_ALLOC_QUICK to avoid blocking on cache * page reuse. */ m = vm_page_alloc(object, mreq->pindex + 1, VM_ALLOC_QUICK); if (m == NULL) { vm_object_drop(object); return(VM_PAGER_OK); } } else { if (m->valid) { vm_page_wakeup(m); vm_object_drop(object); return(VM_PAGER_OK); } vm_page_unqueue_nowakeup(m); } /* page is busy */ mreq = m; raonly = 1; } else { raonly = 0; } /* * Try to block-read contiguous pages from swap if sequential, * otherwise just read one page. Contiguous pages from swap must * reside within a single device stripe because the I/O cannot be * broken up across multiple stripes. * * Note that blk and iblk can be SWAPBLK_NONE but the loop is * set up such that the case(s) are handled implicitly. */ blk = swp_pager_meta_ctl(mreq->object, mreq->pindex, 0); marray[0] = mreq; for (i = 1; i <= swap_burst_read && i < XIO_INTERNAL_PAGES && mreq->pindex + i < object->size; ++i) { swblk_t iblk; iblk = swp_pager_meta_ctl(object, mreq->pindex + i, 0); if (iblk != blk + i) break; if ((blk ^ iblk) & ~SWB_DMMASK) break; m = vm_page_lookup_busy_try(object, mreq->pindex + i, TRUE, &error); if (error) { break; } else if (m == NULL) { /* * Use VM_ALLOC_QUICK to avoid blocking on cache * page reuse. */ m = vm_page_alloc(object, mreq->pindex + i, VM_ALLOC_QUICK); if (m == NULL) break; } else { if (m->valid) { vm_page_wakeup(m); break; } vm_page_unqueue_nowakeup(m); } /* page is busy */ marray[i] = m; } if (i > 1) vm_page_flag_set(marray[i - 1], PG_RAM); /* * If mreq is the requested page and we have nothing to do return * VM_PAGER_FAIL. If raonly is set mreq is just another read-ahead * page and must be cleaned up. */ if (blk == SWAPBLK_NONE) { KKASSERT(i == 1); if (raonly) { vnode_pager_freepage(mreq); vm_object_drop(object); return(VM_PAGER_OK); } else { vm_object_drop(object); return(VM_PAGER_FAIL); } } /* * Map our page(s) into kva for input * * Use the KVABIO API to avoid synchronizing the pmap. */ bp = getpbuf_kva(&nsw_rcount); bio = &bp->b_bio1; kva = (vm_offset_t) bp->b_kvabase; bcopy(marray, bp->b_xio.xio_pages, i * sizeof(vm_page_t)); pmap_qenter_noinval(kva, bp->b_xio.xio_pages, i); bp->b_data = (caddr_t)kva; bp->b_bcount = PAGE_SIZE * i; bp->b_xio.xio_npages = i; bp->b_flags |= B_KVABIO; bio->bio_done = swp_pager_async_iodone; bio->bio_offset = (off_t)blk << PAGE_SHIFT; bio->bio_caller_info1.index = SWBIO_READ; /* * Set index. If raonly set the index beyond the array so all * the pages are treated the same, otherwise the original mreq is * at index 0. */ if (raonly) bio->bio_driver_info = (void *)(intptr_t)i; else bio->bio_driver_info = (void *)(intptr_t)0; for (j = 0; j < i; ++j) { atomic_set_int(&bp->b_xio.xio_pages[j]->busy_count, PBUSY_SWAPINPROG); } mycpu->gd_cnt.v_swapin++; mycpu->gd_cnt.v_swappgsin += bp->b_xio.xio_npages; /* * We still hold the lock on mreq, and our automatic completion routine * does not remove it. */ vm_object_pip_add(object, bp->b_xio.xio_npages); /* * perform the I/O. NOTE!!! bp cannot be considered valid after * this point because we automatically release it on completion. * Instead, we look at the one page we are interested in which we * still hold a lock on even through the I/O completion. * * The other pages in our m[] array are also released on completion, * so we cannot assume they are valid anymore either. */ bp->b_cmd = BUF_CMD_READ; BUF_KERNPROC(bp); vn_strategy(swapdev_vp, bio); /* * Wait for the page we want to complete. PBUSY_SWAPINPROG is always * cleared on completion. If an I/O error occurs, SWAPBLK_NONE * is set in the meta-data. * * If this is a read-ahead only we return immediately without * waiting for I/O. */ if (raonly) { vm_object_drop(object); return(VM_PAGER_OK); } /* * Read-ahead includes originally requested page case. */ for (;;) { busy_count = mreq->busy_count; cpu_ccfence(); if ((busy_count & PBUSY_SWAPINPROG) == 0) break; tsleep_interlock(mreq, 0); if (!atomic_cmpset_int(&mreq->busy_count, busy_count, busy_count | PBUSY_SWAPINPROG | PBUSY_WANTED)) { continue; } atomic_set_int(&mreq->flags, PG_REFERENCED); mycpu->gd_cnt.v_intrans++; if (tsleep(mreq, PINTERLOCKED, "swread", hz*20)) { kprintf( "swap_pager: indefinite wait buffer: " " bp %p offset: %lld, size: %ld " " m=%p busy=%08x flags=%08x\n", bp, (long long)bio->bio_offset, (long)bp->b_bcount, mreq, mreq->busy_count, mreq->flags); } } /* * Disallow speculative reads prior to the SWAPINPROG test. */ cpu_lfence(); /* * mreq is left busied after completion, but all the other pages * are freed. If we had an unrecoverable read error the page will * not be valid. */ vm_object_drop(object); if (mreq->valid != VM_PAGE_BITS_ALL) return(VM_PAGER_ERROR); else return(VM_PAGER_OK); /* * A final note: in a low swap situation, we cannot deallocate swap * and mark a page dirty here because the caller is likely to mark * the page clean when we return, causing the page to possibly revert * to all-zero's later. */ } /* * swap_pager_putpages: * * Assign swap (if necessary) and initiate I/O on the specified pages. * * We support both OBJT_DEFAULT and OBJT_SWAP objects. DEFAULT objects * are automatically converted to SWAP objects. * * In a low memory situation we may block in vn_strategy(), but the new * vm_page reservation system coupled with properly written VFS devices * should ensure that no low-memory deadlock occurs. This is an area * which needs work. * * The parent has N vm_object_pip_add() references prior to * calling us and will remove references for rtvals[] that are * not set to VM_PAGER_PEND. We need to remove the rest on I/O * completion. * * The parent has soft-busy'd the pages it passes us and will unbusy * those whos rtvals[] entry is not set to VM_PAGER_PEND on return. * We need to unbusy the rest on I/O completion. * * No requirements. */ void swap_pager_putpages(vm_object_t object, vm_page_t *m, int count, int flags, int *rtvals) { int i; int n = 0; vm_object_hold(object); if (count && m[0]->object != object) { panic("swap_pager_getpages: object mismatch %p/%p", object, m[0]->object ); } /* * Step 1 * * Turn object into OBJT_SWAP * Check for bogus sysops * * Force sync if not pageout process, we don't want any single * non-pageout process to be able to hog the I/O subsystem! This * can be overridden by setting. */ if (object->type == OBJT_DEFAULT) { if (object->type == OBJT_DEFAULT) swp_pager_meta_convert(object); } /* * Normally we force synchronous swap I/O if this is not the * pageout daemon to prevent any single user process limited * via RLIMIT_RSS from hogging swap write bandwidth. */ if (curthread != pagethread && curthread != emergpager && swap_user_async == 0) { flags |= OBJPC_SYNC; } /* * Step 2 * * Update nsw parameters from swap_async_max sysctl values. * Do not let the sysop crash the machine with bogus numbers. */ if (swap_async_max != nsw_wcount_async_max) { int n; /* * limit range */ if ((n = swap_async_max) > nswbuf_kva / 2) n = nswbuf_kva / 2; if (n < 1) n = 1; swap_async_max = n; /* * Adjust difference ( if possible ). If the current async * count is too low, we may not be able to make the adjustment * at this time. * * vm_token needed for nsw_wcount sleep interlock */ lwkt_gettoken(&vm_token); n -= nsw_wcount_async_max; if (nsw_wcount_async + n >= 0) { nsw_wcount_async_max += n; pbuf_adjcount(&nsw_wcount_async, n); } lwkt_reltoken(&vm_token); } /* * Step 3 * * Assign swap blocks and issue I/O. We reallocate swap on the fly. * The page is left dirty until the pageout operation completes * successfully. */ for (i = 0; i < count; i += n) { struct buf *bp; struct bio *bio; swblk_t blk; int j; /* * Maximum I/O size is limited by a number of factors. */ n = min(BLIST_MAX_ALLOC, count - i); n = min(n, nsw_cluster_max); lwkt_gettoken(&vm_token); /* * Get biggest block of swap we can. If we fail, fall * back and try to allocate a smaller block. Don't go * overboard trying to allocate space if it would overly * fragment swap. */ while ( (blk = swp_pager_getswapspace(object, n)) == SWAPBLK_NONE && n > 4 ) { n >>= 1; } if (blk == SWAPBLK_NONE) { for (j = 0; j < n; ++j) rtvals[i+j] = VM_PAGER_FAIL; lwkt_reltoken(&vm_token); continue; } if (vm_report_swap_allocs > 0) { kprintf("swap_alloc %08jx,%d\n", (intmax_t)blk, n); --vm_report_swap_allocs; } /* * The I/O we are constructing cannot cross a physical * disk boundry in the swap stripe. */ if ((blk ^ (blk + n)) & ~SWB_DMMASK) { j = ((blk + SWB_DMMAX) & ~SWB_DMMASK) - blk; swp_pager_freeswapspace(object, blk + j, n - j); n = j; } /* * All I/O parameters have been satisfied, build the I/O * request and assign the swap space. * * Use the KVABIO API to avoid synchronizing the pmap. */ if ((flags & OBJPC_SYNC)) bp = getpbuf_kva(&nsw_wcount_sync); else bp = getpbuf_kva(&nsw_wcount_async); bio = &bp->b_bio1; lwkt_reltoken(&vm_token); pmap_qenter_noinval((vm_offset_t)bp->b_data, &m[i], n); bp->b_flags |= B_KVABIO; bp->b_bcount = PAGE_SIZE * n; bio->bio_offset = (off_t)blk << PAGE_SHIFT; for (j = 0; j < n; ++j) { vm_page_t mreq = m[i+j]; swp_pager_meta_build(mreq->object, mreq->pindex, blk + j); if (object->type == OBJT_SWAP) vm_page_dirty(mreq); rtvals[i+j] = VM_PAGER_OK; atomic_set_int(&mreq->busy_count, PBUSY_SWAPINPROG); bp->b_xio.xio_pages[j] = mreq; } bp->b_xio.xio_npages = n; mycpu->gd_cnt.v_swapout++; mycpu->gd_cnt.v_swappgsout += bp->b_xio.xio_npages; bp->b_dirtyoff = 0; /* req'd for NFS */ bp->b_dirtyend = bp->b_bcount; /* req'd for NFS */ bp->b_cmd = BUF_CMD_WRITE; bio->bio_caller_info1.index = SWBIO_WRITE; /* * asynchronous */ if ((flags & OBJPC_SYNC) == 0) { bio->bio_done = swp_pager_async_iodone; BUF_KERNPROC(bp); vn_strategy(swapdev_vp, bio); for (j = 0; j < n; ++j) rtvals[i+j] = VM_PAGER_PEND; continue; } /* * Issue synchrnously. * * Wait for the sync I/O to complete, then update rtvals. * We just set the rtvals[] to VM_PAGER_PEND so we can call * our async completion routine at the end, thus avoiding a * double-free. */ bio->bio_caller_info1.index |= SWBIO_SYNC; if (flags & OBJPC_TRY_TO_CACHE) bio->bio_caller_info1.index |= SWBIO_TTC; bio->bio_done = biodone_sync; bio->bio_flags |= BIO_SYNC; vn_strategy(swapdev_vp, bio); biowait(bio, "swwrt"); for (j = 0; j < n; ++j) rtvals[i+j] = VM_PAGER_PEND; /* * Now that we are through with the bp, we can call the * normal async completion, which frees everything up. */ swp_pager_async_iodone(bio); } vm_object_drop(object); } /* * No requirements. * * Recalculate the low and high-water marks. */ void swap_pager_newswap(void) { /* * NOTE: vm_swap_max cannot exceed 1 billion blocks, which is the * limitation imposed by the blist code. Remember that this * will be divided by NSWAP_MAX (4), so each swap device is * limited to around a terrabyte. */ if (vm_swap_max) { nswap_lowat = (int64_t)vm_swap_max * 4 / 100; /* 4% left */ nswap_hiwat = (int64_t)vm_swap_max * 6 / 100; /* 6% left */ kprintf("swap low/high-water marks set to %d/%d\n", nswap_lowat, nswap_hiwat); } else { nswap_lowat = 128; nswap_hiwat = 512; } swp_sizecheck(); } /* * swp_pager_async_iodone: * * Completion routine for asynchronous reads and writes from/to swap. * Also called manually by synchronous code to finish up a bp. * * For READ operations, the pages are BUSY'd. For WRITE operations, * the pages are vm_page_t->busy'd. For READ operations, we BUSY * unbusy all pages except the 'main' request page. For WRITE * operations, we vm_page_t->busy'd unbusy all pages ( we can do this * because we marked them all VM_PAGER_PEND on return from putpages ). * * This routine may not block. * * No requirements. */ static void swp_pager_async_iodone(struct bio *bio) { struct buf *bp = bio->bio_buf; vm_object_t object = NULL; int i; int *nswptr; /* * report error */ if (bp->b_flags & B_ERROR) { kprintf( "swap_pager: I/O error - %s failed; offset %lld," "size %ld, error %d\n", ((bio->bio_caller_info1.index & SWBIO_READ) ? "pagein" : "pageout"), (long long)bio->bio_offset, (long)bp->b_bcount, bp->b_error ); } /* * set object. */ if (bp->b_xio.xio_npages) object = bp->b_xio.xio_pages[0]->object; #if 0 /* PMAP TESTING CODE (useful, keep it in but #if 0'd) */ if (bio->bio_caller_info1.index & SWBIO_WRITE) { if (bio->bio_crc != iscsi_crc32(bp->b_data, bp->b_bcount)) { kprintf("SWAPOUT: BADCRC %08x %08x\n", bio->bio_crc, iscsi_crc32(bp->b_data, bp->b_bcount)); for (i = 0; i < bp->b_xio.xio_npages; ++i) { vm_page_t m = bp->b_xio.xio_pages[i]; if ((m->flags & PG_WRITEABLE) && (pmap_mapped_sync(m) & PG_WRITEABLE)) { kprintf("SWAPOUT: " "%d/%d %p writable\n", i, bp->b_xio.xio_npages, m); } } } } #endif /* * remove the mapping for kernel virtual */ pmap_qremove((vm_offset_t)bp->b_data, bp->b_xio.xio_npages); /* * cleanup pages. If an error occurs writing to swap, we are in * very serious trouble. If it happens to be a disk error, though, * we may be able to recover by reassigning the swap later on. So * in this case we remove the m->swapblk assignment for the page * but do not free it in the rlist. The errornous block(s) are thus * never reallocated as swap. Redirty the page and continue. */ for (i = 0; i < bp->b_xio.xio_npages; ++i) { vm_page_t m = bp->b_xio.xio_pages[i]; if (bp->b_flags & B_ERROR) { /* * If an error occurs I'd love to throw the swapblk * away without freeing it back to swapspace, so it * can never be used again. But I can't from an * interrupt. */ if (bio->bio_caller_info1.index & SWBIO_READ) { /* * When reading, reqpage needs to stay * locked for the parent, but all other * pages can be freed. We still want to * wakeup the parent waiting on the page, * though. ( also: pg_reqpage can be -1 and * not match anything ). * * We have to wake specifically requested pages * up too because we cleared SWAPINPROG and * someone may be waiting for that. * * NOTE: For reads, m->dirty will probably * be overridden by the original caller * of getpages so don't play cute tricks * here. * * NOTE: We can't actually free the page from * here, because this is an interrupt. * It is not legal to mess with * object->memq from an interrupt. * Deactivate the page instead. * * WARNING! The instant SWAPINPROG is * cleared another cpu may start * using the mreq page (it will * check m->valid immediately). */ m->valid = 0; atomic_clear_int(&m->busy_count, PBUSY_SWAPINPROG); /* * bio_driver_info holds the requested page * index. */ if (i != (int)(intptr_t)bio->bio_driver_info) { vm_page_deactivate(m); vm_page_wakeup(m); } else { vm_page_flash(m); } /* * If i == bp->b_pager.pg_reqpage, do not wake * the page up. The caller needs to. */ } else { /* * If a write error occurs remove the swap * assignment (note that PG_SWAPPED may or * may not be set depending on prior activity). * * Re-dirty OBJT_SWAP pages as there is no * other backing store, we can't throw the * page away. * * Non-OBJT_SWAP pages (aka swapcache) must * not be dirtied since they may not have * been dirty in the first place, and they * do have backing store (the vnode). */ vm_page_busy_wait(m, FALSE, "swadpg"); vm_object_hold(m->object); swp_pager_meta_ctl(m->object, m->pindex, SWM_FREE); vm_page_flag_clear(m, PG_SWAPPED); vm_object_drop(m->object); if (m->object->type == OBJT_SWAP) { vm_page_dirty(m); vm_page_activate(m); } vm_page_io_finish(m); atomic_clear_int(&m->busy_count, PBUSY_SWAPINPROG); vm_page_wakeup(m); } } else if (bio->bio_caller_info1.index & SWBIO_READ) { /* * NOTE: for reads, m->dirty will probably be * overridden by the original caller of getpages so * we cannot set them in order to free the underlying * swap in a low-swap situation. I don't think we'd * want to do that anyway, but it was an optimization * that existed in the old swapper for a time before * it got ripped out due to precisely this problem. * * If not the requested page then deactivate it. * * Note that the requested page, reqpage, is left * busied, but we still have to wake it up. The * other pages are released (unbusied) by * vm_page_wakeup(). We do not set reqpage's * valid bits here, it is up to the caller. */ /* * NOTE: Can't call pmap_clear_modify(m) from an * interrupt thread, the pmap code may have to * map non-kernel pmaps and currently asserts * the case. * * WARNING! The instant SWAPINPROG is * cleared another cpu may start * using the mreq page (it will * check m->valid immediately). */ /*pmap_clear_modify(m);*/ m->valid = VM_PAGE_BITS_ALL; vm_page_undirty(m); vm_page_flag_set(m, PG_SWAPPED); atomic_clear_int(&m->busy_count, PBUSY_SWAPINPROG); /* * We have to wake specifically requested pages * up too because we cleared SWAPINPROG and * could be waiting for it in getpages. However, * be sure to not unbusy getpages specifically * requested page - getpages expects it to be * left busy. * * bio_driver_info holds the requested page */ if (i != (int)(intptr_t)bio->bio_driver_info) { vm_page_deactivate(m); vm_page_wakeup(m); } else { vm_page_flash(m); } } else { /* * Mark the page clean but do not mess with the * pmap-layer's modified state. That state should * also be clear since the caller protected the * page VM_PROT_READ, but allow the case. * * We are in an interrupt, avoid pmap operations. * * If we have a severe page deficit, deactivate the * page. Do not try to cache it (which would also * involve a pmap op), because the page might still * be read-heavy. * * When using the swap to cache clean vnode pages * we do not mess with the page dirty bits. * * NOTE! Nobody is waiting for the key mreq page * on write completion. */ vm_page_busy_wait(m, FALSE, "swadpg"); if (m->object->type == OBJT_SWAP) vm_page_undirty(m); vm_page_flag_set(m, PG_SWAPPED); atomic_clear_int(&m->busy_count, PBUSY_SWAPINPROG); if (vm_paging_severe()) vm_page_deactivate(m); vm_page_io_finish(m); if (bio->bio_caller_info1.index & SWBIO_TTC) vm_page_try_to_cache(m); else vm_page_wakeup(m); } } /* * adjust pip. NOTE: the original parent may still have its own * pip refs on the object. */ if (object) vm_object_pip_wakeup_n(object, bp->b_xio.xio_npages); /* * Release the physical I/O buffer. * * NOTE: Due to synchronous operations in the write case b_cmd may * already be set to BUF_CMD_DONE and BIO_SYNC may have already * been cleared. * * Use vm_token to interlock nsw_rcount/wcount wakeup? */ lwkt_gettoken(&vm_token); if (bio->bio_caller_info1.index & SWBIO_READ) nswptr = &nsw_rcount; else if (bio->bio_caller_info1.index & SWBIO_SYNC) nswptr = &nsw_wcount_sync; else nswptr = &nsw_wcount_async; bp->b_cmd = BUF_CMD_DONE; relpbuf(bp, nswptr); lwkt_reltoken(&vm_token); } /* * Fault-in a potentially swapped page and remove the swap reference. * (used by swapoff code) * * object must be held. */ static __inline void swp_pager_fault_page(vm_object_t object, int *sharedp, vm_pindex_t pindex) { struct vnode *vp; vm_page_t m; int error; ASSERT_LWKT_TOKEN_HELD(vm_object_token(object)); if (object->type == OBJT_VNODE) { /* * Any swap related to a vnode is due to swapcache. We must * vget() the vnode in case it is not active (otherwise * vref() will panic). Calling vm_object_page_remove() will * ensure that any swap ref is removed interlocked with the * page. clean_only is set to TRUE so we don't throw away * dirty pages. */ vp = object->handle; error = vget(vp, LK_SHARED | LK_RETRY | LK_CANRECURSE); if (error == 0) { vm_object_page_remove(object, pindex, pindex + 1, TRUE); vput(vp); } } else { /* * Otherwise it is a normal OBJT_SWAP object and we can * fault the page in and remove the swap. */ m = vm_fault_object_page(object, IDX_TO_OFF(pindex), VM_PROT_NONE, VM_FAULT_DIRTY | VM_FAULT_UNSWAP, sharedp, &error); if (m) vm_page_unhold(m); } } /* * This removes all swap blocks related to a particular device. We have * to be careful of ripups during the scan. */ static int swp_pager_swapoff_callback(struct swblock *swap, void *data); int swap_pager_swapoff(int devidx) { struct vm_object_hash *hash; struct swswapoffinfo info; struct vm_object marker; vm_object_t object; int n; bzero(&marker, sizeof(marker)); marker.type = OBJT_MARKER; for (n = 0; n < VMOBJ_HSIZE; ++n) { hash = &vm_object_hash[n]; lwkt_gettoken(&hash->token); TAILQ_INSERT_HEAD(&hash->list, &marker, object_entry); while ((object = TAILQ_NEXT(&marker, object_entry)) != NULL) { if (object->type == OBJT_MARKER) goto skip; if (object->type != OBJT_SWAP && object->type != OBJT_VNODE) goto skip; vm_object_hold(object); if (object->type != OBJT_SWAP && object->type != OBJT_VNODE) { vm_object_drop(object); goto skip; } /* * Object is special in that we can't just pagein * into vm_page's in it (tmpfs, vn). */ if ((object->flags & OBJ_NOPAGEIN) && RB_ROOT(&object->swblock_root)) { vm_object_drop(object); goto skip; } info.object = object; info.shared = 0; info.devidx = devidx; swblock_rb_tree_RB_SCAN(&object->swblock_root, NULL, swp_pager_swapoff_callback, &info); vm_object_drop(object); skip: if (object == TAILQ_NEXT(&marker, object_entry)) { TAILQ_REMOVE(&hash->list, &marker, object_entry); TAILQ_INSERT_AFTER(&hash->list, object, &marker, object_entry); } } TAILQ_REMOVE(&hash->list, &marker, object_entry); lwkt_reltoken(&hash->token); } /* * If we fail to locate all swblocks we just fail gracefully and * do not bother to restore paging on the swap device. If the * user wants to retry the user can retry. */ if (swdevt[devidx].sw_nused) return (1); else return (0); } static int swp_pager_swapoff_callback(struct swblock *swap, void *data) { struct swswapoffinfo *info = data; vm_object_t object = info->object; vm_pindex_t index; swblk_t v; int i; index = swap->swb_index; for (i = 0; i < SWAP_META_PAGES; ++i) { /* * Make sure we don't race a dying object. This will * kill the scan of the object's swap blocks entirely. */ if (object->flags & OBJ_DEAD) return(-1); /* * Fault the page, which can obviously block. If the swap * structure disappears break out. */ v = swap->swb_pages[i]; if (v != SWAPBLK_NONE && BLK2DEVIDX(v) == info->devidx) { swp_pager_fault_page(object, &info->shared, swap->swb_index + i); /* swap ptr might go away */ if (RB_LOOKUP(swblock_rb_tree, &object->swblock_root, index) != swap) { break; } } } return(0); } /************************************************************************ * SWAP META DATA * ************************************************************************ * * These routines manipulate the swap metadata stored in the * OBJT_SWAP object. * * Swap metadata is implemented with a global hash and not directly * linked into the object. Instead the object simply contains * appropriate tracking counters. */ /* * Lookup the swblock containing the specified swap block index. * * The caller must hold the object. */ static __inline struct swblock * swp_pager_lookup(vm_object_t object, vm_pindex_t index) { ASSERT_LWKT_TOKEN_HELD(vm_object_token(object)); index &= ~(vm_pindex_t)SWAP_META_MASK; return (RB_LOOKUP(swblock_rb_tree, &object->swblock_root, index)); } /* * Remove a swblock from the RB tree. * * The caller must hold the object. */ static __inline void swp_pager_remove(vm_object_t object, struct swblock *swap) { ASSERT_LWKT_TOKEN_HELD(vm_object_token(object)); RB_REMOVE(swblock_rb_tree, &object->swblock_root, swap); } /* * Convert default object to swap object if necessary * * The caller must hold the object. */ static void swp_pager_meta_convert(vm_object_t object) { if (object->type == OBJT_DEFAULT) { object->type = OBJT_SWAP; KKASSERT(object->swblock_count == 0); } } /* * SWP_PAGER_META_BUILD() - add swap block to swap meta data for object * * We first convert the object to a swap object if it is a default * object. Vnode objects do not need to be converted. * * The specified swapblk is added to the object's swap metadata. If * the swapblk is not valid, it is freed instead. Any previously * assigned swapblk is freed. * * The caller must hold the object. */ static void swp_pager_meta_build(vm_object_t object, vm_pindex_t index, swblk_t swapblk) { struct swblock *swap; struct swblock *oswap; vm_pindex_t v; KKASSERT(swapblk != SWAPBLK_NONE); ASSERT_LWKT_TOKEN_HELD(vm_object_token(object)); /* * Convert object if necessary */ if (object->type == OBJT_DEFAULT) swp_pager_meta_convert(object); /* * Locate swblock. If not found create, but if we aren't adding * anything just return. If we run out of space in the map we wait * and, since the hash table may have changed, retry. */ retry: swap = swp_pager_lookup(object, index); if (swap == NULL) { int i; swap = zalloc(swap_zone); if (swap == NULL) { vm_wait(0); goto retry; } swap->swb_index = index & ~(vm_pindex_t)SWAP_META_MASK; swap->swb_count = 0; ++object->swblock_count; for (i = 0; i < SWAP_META_PAGES; ++i) swap->swb_pages[i] = SWAPBLK_NONE; oswap = RB_INSERT(swblock_rb_tree, &object->swblock_root, swap); KKASSERT(oswap == NULL); } /* * Delete prior contents of metadata. * * NOTE: Decrement swb_count after the freeing operation (which * might block) to prevent racing destruction of the swblock. */ index &= SWAP_META_MASK; while ((v = swap->swb_pages[index]) != SWAPBLK_NONE) { swap->swb_pages[index] = SWAPBLK_NONE; /* can block */ swp_pager_freeswapspace(object, v, 1); --swap->swb_count; --mycpu->gd_vmtotal.t_vm; } /* * Enter block into metadata */ swap->swb_pages[index] = swapblk; if (swapblk != SWAPBLK_NONE) { ++swap->swb_count; ++mycpu->gd_vmtotal.t_vm; } } /* * SWP_PAGER_META_FREE() - free a range of blocks in the object's swap metadata * * The requested range of blocks is freed, with any associated swap * returned to the swap bitmap. * * This routine will free swap metadata structures as they are cleaned * out. This routine does *NOT* operate on swap metadata associated * with resident pages. * * The caller must hold the object. */ static int swp_pager_meta_free_callback(struct swblock *swb, void *data); static void swp_pager_meta_free(vm_object_t object, vm_pindex_t index, vm_pindex_t count) { struct swfreeinfo info; ASSERT_LWKT_TOKEN_HELD(vm_object_token(object)); /* * Nothing to do */ if (object->swblock_count == 0) { KKASSERT(RB_EMPTY(&object->swblock_root)); return; } if (count == 0) return; /* * Setup for RB tree scan. Note that the pindex range can be huge * due to the 64 bit page index space so we cannot safely iterate. */ info.object = object; info.basei = index & ~(vm_pindex_t)SWAP_META_MASK; info.begi = index; info.endi = index + count - 1; swblock_rb_tree_RB_SCAN(&object->swblock_root, rb_swblock_scancmp, swp_pager_meta_free_callback, &info); } /* * The caller must hold the object. */ static int swp_pager_meta_free_callback(struct swblock *swap, void *data) { struct swfreeinfo *info = data; vm_object_t object = info->object; int index; int eindex; /* * Figure out the range within the swblock. The wider scan may * return edge-case swap blocks when the start and/or end points * are in the middle of a block. */ if (swap->swb_index < info->begi) index = (int)info->begi & SWAP_META_MASK; else index = 0; if (swap->swb_index + SWAP_META_PAGES > info->endi) eindex = (int)info->endi & SWAP_META_MASK; else eindex = SWAP_META_MASK; /* * Scan and free the blocks. The loop terminates early * if (swap) runs out of blocks and could be freed. * * NOTE: Decrement swb_count after swp_pager_freeswapspace() * to deal with a zfree race. */ while (index <= eindex) { swblk_t v = swap->swb_pages[index]; if (v != SWAPBLK_NONE) { swap->swb_pages[index] = SWAPBLK_NONE; /* can block */ swp_pager_freeswapspace(object, v, 1); --mycpu->gd_vmtotal.t_vm; if (--swap->swb_count == 0) { swp_pager_remove(object, swap); zfree(swap_zone, swap); --object->swblock_count; break; } } ++index; } /* swap may be invalid here due to zfree above */ lwkt_yield(); return(0); } /* * SWP_PAGER_META_FREE_ALL() - destroy all swap metadata associated with object * * This routine locates and destroys all swap metadata associated with * an object. * * NOTE: Decrement swb_count after the freeing operation (which * might block) to prevent racing destruction of the swblock. * * The caller must hold the object. */ static void swp_pager_meta_free_all(vm_object_t object) { struct swblock *swap; int i; ASSERT_LWKT_TOKEN_HELD(vm_object_token(object)); while ((swap = RB_ROOT(&object->swblock_root)) != NULL) { swp_pager_remove(object, swap); for (i = 0; i < SWAP_META_PAGES; ++i) { swblk_t v = swap->swb_pages[i]; if (v != SWAPBLK_NONE) { /* can block */ swp_pager_freeswapspace(object, v, 1); --swap->swb_count; --mycpu->gd_vmtotal.t_vm; } } if (swap->swb_count != 0) panic("swap_pager_meta_free_all: swb_count != 0"); zfree(swap_zone, swap); --object->swblock_count; lwkt_yield(); } KKASSERT(object->swblock_count == 0); } /* * SWP_PAGER_METACTL() - misc control of swap and vm_page_t meta data. * * This routine is capable of looking up, popping, or freeing * swapblk assignments in the swap meta data or in the vm_page_t. * The routine typically returns the swapblk being looked-up, or popped, * or SWAPBLK_NONE if the block was freed, or SWAPBLK_NONE if the block * was invalid. This routine will automatically free any invalid * meta-data swapblks. * * It is not possible to store invalid swapblks in the swap meta data * (other then a literal 'SWAPBLK_NONE'), so we don't bother checking. * * When acting on a busy resident page and paging is in progress, we * have to wait until paging is complete but otherwise can act on the * busy page. * * SWM_FREE remove and free swap block from metadata * SWM_POP remove from meta data but do not free.. pop it out * * The caller must hold the object. */ static swblk_t swp_pager_meta_ctl(vm_object_t object, vm_pindex_t index, int flags) { struct swblock *swap; swblk_t r1; if (object->swblock_count == 0) return(SWAPBLK_NONE); r1 = SWAPBLK_NONE; swap = swp_pager_lookup(object, index); if (swap != NULL) { index &= SWAP_META_MASK; r1 = swap->swb_pages[index]; if (r1 != SWAPBLK_NONE) { if (flags & (SWM_FREE|SWM_POP)) { swap->swb_pages[index] = SWAPBLK_NONE; --mycpu->gd_vmtotal.t_vm; if (--swap->swb_count == 0) { swp_pager_remove(object, swap); zfree(swap_zone, swap); --object->swblock_count; } } /* swap ptr may be invalid */ if (flags & SWM_FREE) { swp_pager_freeswapspace(object, r1, 1); r1 = SWAPBLK_NONE; } } /* swap ptr may be invalid */ } return(r1); }
