Index: head/lib/libc/db/hash/hash.h =================================================================== --- head/lib/libc/db/hash/hash.h (revision 295030) +++ head/lib/libc/db/hash/hash.h (revision 295031) @@ -1,290 +1,290 @@ /*- * Copyright (c) 1990, 1993, 1994 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Margo Seltzer. * * 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. * 4. 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. * * @(#)hash.h 8.3 (Berkeley) 5/31/94 * $FreeBSD$ */ /* Operations */ typedef enum { HASH_GET, HASH_PUT, HASH_PUTNEW, HASH_DELETE, HASH_FIRST, HASH_NEXT } ACTION; /* Buffer Management structures */ typedef struct _bufhead BUFHEAD; struct _bufhead { BUFHEAD *prev; /* LRU links */ BUFHEAD *next; /* LRU links */ BUFHEAD *ovfl; /* Overflow page buffer header */ u_int32_t addr; /* Address of this page */ char *page; /* Actual page data */ char flags; #define BUF_MOD 0x0001 #define BUF_DISK 0x0002 #define BUF_BUCKET 0x0004 #define BUF_PIN 0x0008 }; #define IS_BUCKET(X) ((X) & BUF_BUCKET) typedef BUFHEAD **SEGMENT; /* Hash Table Information */ typedef struct hashhdr { /* Disk resident portion */ int32_t magic; /* Magic NO for hash tables */ int32_t version; /* Version ID */ u_int32_t lorder; /* Byte Order */ int32_t bsize; /* Bucket/Page Size */ int32_t bshift; /* Bucket shift */ int32_t dsize; /* Directory Size */ int32_t ssize; /* Segment Size */ int32_t sshift; /* Segment shift */ int32_t ovfl_point; /* Where overflow pages are being * allocated */ int32_t last_freed; /* Last overflow page freed */ u_int32_t max_bucket; /* ID of Maximum bucket in use */ u_int32_t high_mask; /* Mask to modulo into entire table */ u_int32_t low_mask; /* Mask to modulo into lower half of * table */ u_int32_t ffactor; /* Fill factor */ int32_t nkeys; /* Number of keys in hash table */ int32_t hdrpages; /* Size of table header */ int32_t h_charkey; /* value of hash(CHARKEY) */ #define NCACHED 32 /* number of bit maps and spare * points */ int32_t spares[NCACHED];/* spare pages for overflow */ u_int16_t bitmaps[NCACHED]; /* address of overflow page * bitmaps */ } HASHHDR; typedef struct htab { /* Memory resident data structure */ HASHHDR hdr; /* Header */ int nsegs; /* Number of allocated segments */ int exsegs; /* Number of extra allocated * segments */ u_int32_t /* Hash function */ (*hash)(const void *, size_t); int flags; /* Flag values */ int fp; /* File pointer */ char *tmp_buf; /* Temporary Buffer for BIG data */ char *tmp_key; /* Temporary Buffer for BIG keys */ BUFHEAD *cpage; /* Current page */ int cbucket; /* Current bucket */ int cndx; /* Index of next item on cpage */ int error; /* Error Number -- for DBM * compatibility */ int new_file; /* Indicates if fd is backing store * or no */ int save_file; /* Indicates whether we need to flush * file at * exit */ u_int32_t *mapp[NCACHED]; /* Pointers to page maps */ int nmaps; /* Initial number of bitmaps */ int nbufs; /* Number of buffers left to * allocate */ BUFHEAD bufhead; /* Header of buffer lru list */ SEGMENT *dir; /* Hash Bucket directory */ } HTAB; /* * Constants */ #define MAX_BSIZE 32768 /* 2^15 but should be 65536 */ #define MIN_BUFFERS 6 #define MINHDRSIZE 512 #define DEF_BUFSIZE 65536 /* 64 K */ #define DEF_BUCKET_SIZE 4096 #define DEF_BUCKET_SHIFT 12 /* log2(BUCKET) */ #define DEF_SEGSIZE 256 #define DEF_SEGSIZE_SHIFT 8 /* log2(SEGSIZE) */ #define DEF_DIRSIZE 256 #define DEF_FFACTOR 65536 #define MIN_FFACTOR 4 #define SPLTMAX 8 #define CHARKEY "%$sniglet^&" #define NUMKEY 1038583 #define BYTE_SHIFT 3 #define INT_TO_BYTE 2 #define INT_BYTE_SHIFT 5 #define ALL_SET ((u_int32_t)0xFFFFFFFF) #define ALL_CLEAR 0 -#define PTROF(X) ((BUFHEAD *)((ptrdiff_t)(X)&~0x3)) -#define ISMOD(X) ((u_int32_t)(ptrdiff_t)(X)&0x1) -#define DOMOD(X) ((X) = (char *)((ptrdiff_t)(X)|0x1)) -#define ISDISK(X) ((u_int32_t)(ptrdiff_t)(X)&0x2) -#define DODISK(X) ((X) = (char *)((ptrdiff_t)(X)|0x2)) +#define PTROF(X) ((BUFHEAD *)((intptr_t)(X)&~0x3)) +#define ISMOD(X) ((u_int32_t)(intptr_t)(X)&0x1) +#define DOMOD(X) ((X) = (char *)((intptr_t)(X)|0x1)) +#define ISDISK(X) ((u_int32_t)(intptr_t)(X)&0x2) +#define DODISK(X) ((X) = (char *)((intptr_t)(X)|0x2)) #define BITS_PER_MAP 32 /* Given the address of the beginning of a big map, clear/set the nth bit */ #define CLRBIT(A, N) ((A)[(N)/BITS_PER_MAP] &= ~(1<<((N)%BITS_PER_MAP))) #define SETBIT(A, N) ((A)[(N)/BITS_PER_MAP] |= (1<<((N)%BITS_PER_MAP))) #define ISSET(A, N) ((A)[(N)/BITS_PER_MAP] & (1<<((N)%BITS_PER_MAP))) /* Overflow management */ /* * Overflow page numbers are allocated per split point. At each doubling of * the table, we can allocate extra pages. So, an overflow page number has * the top 5 bits indicate which split point and the lower 11 bits indicate * which page at that split point is indicated (pages within split points are * numberered starting with 1). */ #define SPLITSHIFT 11 #define SPLITMASK 0x7FF #define SPLITNUM(N) (((u_int32_t)(N)) >> SPLITSHIFT) #define OPAGENUM(N) ((N) & SPLITMASK) #define OADDR_OF(S,O) ((u_int32_t)((u_int32_t)(S) << SPLITSHIFT) + (O)) #define BUCKET_TO_PAGE(B) \ (B) + hashp->HDRPAGES + ((B) ? hashp->SPARES[__log2((B)+1)-1] : 0) #define OADDR_TO_PAGE(B) \ BUCKET_TO_PAGE ( (1 << SPLITNUM((B))) -1 ) + OPAGENUM((B)); /* * page.h contains a detailed description of the page format. * * Normally, keys and data are accessed from offset tables in the top of * each page which point to the beginning of the key and data. There are * four flag values which may be stored in these offset tables which indicate * the following: * * * OVFLPAGE Rather than a key data pair, this pair contains * the address of an overflow page. The format of * the pair is: * OVERFLOW_PAGE_NUMBER OVFLPAGE * * PARTIAL_KEY This must be the first key/data pair on a page * and implies that page contains only a partial key. * That is, the key is too big to fit on a single page * so it starts on this page and continues on the next. * The format of the page is: * KEY_OFF PARTIAL_KEY OVFL_PAGENO OVFLPAGE * * KEY_OFF -- offset of the beginning of the key * PARTIAL_KEY -- 1 * OVFL_PAGENO - page number of the next overflow page * OVFLPAGE -- 0 * * FULL_KEY This must be the first key/data pair on the page. It * is used in two cases. * * Case 1: * There is a complete key on the page but no data * (because it wouldn't fit). The next page contains * the data. * * Page format it: * KEY_OFF FULL_KEY OVFL_PAGENO OVFL_PAGE * * KEY_OFF -- offset of the beginning of the key * FULL_KEY -- 2 * OVFL_PAGENO - page number of the next overflow page * OVFLPAGE -- 0 * * Case 2: * This page contains no key, but part of a large * data field, which is continued on the next page. * * Page format it: * DATA_OFF FULL_KEY OVFL_PAGENO OVFL_PAGE * * KEY_OFF -- offset of the beginning of the data on * this page * FULL_KEY -- 2 * OVFL_PAGENO - page number of the next overflow page * OVFLPAGE -- 0 * * FULL_KEY_DATA * This must be the first key/data pair on the page. * There are two cases: * * Case 1: * This page contains a key and the beginning of the * data field, but the data field is continued on the * next page. * * Page format is: * KEY_OFF FULL_KEY_DATA OVFL_PAGENO DATA_OFF * * KEY_OFF -- offset of the beginning of the key * FULL_KEY_DATA -- 3 * OVFL_PAGENO - page number of the next overflow page * DATA_OFF -- offset of the beginning of the data * * Case 2: * This page contains the last page of a big data pair. * There is no key, only the tail end of the data * on this page. * * Page format is: * DATA_OFF FULL_KEY_DATA * * DATA_OFF -- offset of the beginning of the data on * this page * FULL_KEY_DATA -- 3 * OVFL_PAGENO - page number of the next overflow page * OVFLPAGE -- 0 * * OVFL_PAGENO and OVFLPAGE are optional (they are * not present if there is no next page). */ #define OVFLPAGE 0 #define PARTIAL_KEY 1 #define FULL_KEY 2 #define FULL_KEY_DATA 3 #define REAL_KEY 4 /* Short hands for accessing structure */ #define BSIZE hdr.bsize #define BSHIFT hdr.bshift #define DSIZE hdr.dsize #define SGSIZE hdr.ssize #define SSHIFT hdr.sshift #define LORDER hdr.lorder #define OVFL_POINT hdr.ovfl_point #define LAST_FREED hdr.last_freed #define MAX_BUCKET hdr.max_bucket #define FFACTOR hdr.ffactor #define HIGH_MASK hdr.high_mask #define LOW_MASK hdr.low_mask #define NKEYS hdr.nkeys #define HDRPAGES hdr.hdrpages #define SPARES hdr.spares #define BITMAPS hdr.bitmaps #define VERSION hdr.version #define MAGIC hdr.magic #define NEXT_FREE hdr.next_free #define H_CHARKEY hdr.h_charkey Index: head/lib/libc/db/hash/hash_buf.c =================================================================== --- head/lib/libc/db/hash/hash_buf.c (revision 295030) +++ head/lib/libc/db/hash/hash_buf.c (revision 295031) @@ -1,358 +1,358 @@ /*- * Copyright (c) 1990, 1993, 1994 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Margo Seltzer. * * 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. * 4. 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. */ #if defined(LIBC_SCCS) && !defined(lint) static char sccsid[] = "@(#)hash_buf.c 8.5 (Berkeley) 7/15/94"; #endif /* LIBC_SCCS and not lint */ #include __FBSDID("$FreeBSD$"); /* * PACKAGE: hash * * DESCRIPTION: * Contains buffer management * * ROUTINES: * External * __buf_init * __get_buf * __buf_free * __reclaim_buf * Internal * newbuf */ #include #include #include #include #include #ifdef DEBUG #include #endif #include #include "hash.h" #include "page.h" #include "extern.h" static BUFHEAD *newbuf(HTAB *, u_int32_t, BUFHEAD *); /* Unlink B from its place in the lru */ #define BUF_REMOVE(B) { \ (B)->prev->next = (B)->next; \ (B)->next->prev = (B)->prev; \ } /* Insert B after P */ #define BUF_INSERT(B, P) { \ (B)->next = (P)->next; \ (B)->prev = (P); \ (P)->next = (B); \ (B)->next->prev = (B); \ } #define MRU hashp->bufhead.next #define LRU hashp->bufhead.prev #define MRU_INSERT(B) BUF_INSERT((B), &hashp->bufhead) #define LRU_INSERT(B) BUF_INSERT((B), LRU) /* * We are looking for a buffer with address "addr". If prev_bp is NULL, then * address is a bucket index. If prev_bp is not NULL, then it points to the * page previous to an overflow page that we are trying to find. * * CAVEAT: The buffer header accessed via prev_bp's ovfl field may no longer * be valid. Therefore, you must always verify that its address matches the * address you are seeking. */ BUFHEAD * __get_buf(HTAB *hashp, u_int32_t addr, BUFHEAD *prev_bp, /* If prev_bp set, indicates a new overflow page. */ int newpage) { BUFHEAD *bp; u_int32_t is_disk_mask; int is_disk, segment_ndx; SEGMENT segp; is_disk = 0; is_disk_mask = 0; if (prev_bp) { bp = prev_bp->ovfl; if (!bp || (bp->addr != addr)) bp = NULL; if (!newpage) is_disk = BUF_DISK; } else { /* Grab buffer out of directory */ segment_ndx = addr & (hashp->SGSIZE - 1); /* valid segment ensured by __call_hash() */ segp = hashp->dir[addr >> hashp->SSHIFT]; #ifdef DEBUG assert(segp != NULL); #endif bp = PTROF(segp[segment_ndx]); is_disk_mask = ISDISK(segp[segment_ndx]); is_disk = is_disk_mask || !hashp->new_file; } if (!bp) { bp = newbuf(hashp, addr, prev_bp); if (!bp || __get_page(hashp, bp->page, addr, !prev_bp, is_disk, 0)) return (NULL); if (!prev_bp) segp[segment_ndx] = - (BUFHEAD *)((ptrdiff_t)bp | is_disk_mask); + (BUFHEAD *)((intptr_t)bp | is_disk_mask); } else { BUF_REMOVE(bp); MRU_INSERT(bp); } return (bp); } /* * We need a buffer for this page. Either allocate one, or evict a resident * one (if we have as many buffers as we're allowed) and put this one in. * * If newbuf finds an error (returning NULL), it also sets errno. */ static BUFHEAD * newbuf(HTAB *hashp, u_int32_t addr, BUFHEAD *prev_bp) { BUFHEAD *bp; /* The buffer we're going to use */ BUFHEAD *xbp; /* Temp pointer */ BUFHEAD *next_xbp; SEGMENT segp; int segment_ndx; u_int16_t oaddr, *shortp; oaddr = 0; bp = LRU; /* It is bad to overwrite the page under the cursor. */ if (bp == hashp->cpage) { BUF_REMOVE(bp); MRU_INSERT(bp); bp = LRU; } /* If prev_bp is part of bp overflow, create a new buffer. */ if (hashp->nbufs == 0 && prev_bp && bp->ovfl) { BUFHEAD *ovfl; for (ovfl = bp->ovfl; ovfl ; ovfl = ovfl->ovfl) { if (ovfl == prev_bp) { hashp->nbufs++; break; } } } /* * If LRU buffer is pinned, the buffer pool is too small. We need to * allocate more buffers. */ if (hashp->nbufs || (bp->flags & BUF_PIN) || bp == hashp->cpage) { /* Allocate a new one */ if ((bp = (BUFHEAD *)calloc(1, sizeof(BUFHEAD))) == NULL) return (NULL); if ((bp->page = (char *)calloc(1, hashp->BSIZE)) == NULL) { free(bp); return (NULL); } if (hashp->nbufs) hashp->nbufs--; } else { /* Kick someone out */ BUF_REMOVE(bp); /* * If this is an overflow page with addr 0, it's already been * flushed back in an overflow chain and initialized. */ if ((bp->addr != 0) || (bp->flags & BUF_BUCKET)) { /* * Set oaddr before __put_page so that you get it * before bytes are swapped. */ shortp = (u_int16_t *)bp->page; if (shortp[0]) oaddr = shortp[shortp[0] - 1]; if ((bp->flags & BUF_MOD) && __put_page(hashp, bp->page, bp->addr, (int)IS_BUCKET(bp->flags), 0)) return (NULL); /* * Update the pointer to this page (i.e. invalidate it). * * If this is a new file (i.e. we created it at open * time), make sure that we mark pages which have been * written to disk so we retrieve them from disk later, * rather than allocating new pages. */ if (IS_BUCKET(bp->flags)) { segment_ndx = bp->addr & (hashp->SGSIZE - 1); segp = hashp->dir[bp->addr >> hashp->SSHIFT]; #ifdef DEBUG assert(segp != NULL); #endif if (hashp->new_file && ((bp->flags & BUF_MOD) || ISDISK(segp[segment_ndx]))) segp[segment_ndx] = (BUFHEAD *)BUF_DISK; else segp[segment_ndx] = NULL; } /* * Since overflow pages can only be access by means of * their bucket, free overflow pages associated with * this bucket. */ for (xbp = bp; xbp->ovfl;) { next_xbp = xbp->ovfl; xbp->ovfl = 0; xbp = next_xbp; /* Check that ovfl pointer is up date. */ if (IS_BUCKET(xbp->flags) || (oaddr != xbp->addr)) break; shortp = (u_int16_t *)xbp->page; if (shortp[0]) /* set before __put_page */ oaddr = shortp[shortp[0] - 1]; if ((xbp->flags & BUF_MOD) && __put_page(hashp, xbp->page, xbp->addr, 0, 0)) return (NULL); xbp->addr = 0; xbp->flags = 0; BUF_REMOVE(xbp); LRU_INSERT(xbp); } } } /* Now assign this buffer */ bp->addr = addr; #ifdef DEBUG1 (void)fprintf(stderr, "NEWBUF1: %d->ovfl was %d is now %d\n", bp->addr, (bp->ovfl ? bp->ovfl->addr : 0), 0); #endif bp->ovfl = NULL; if (prev_bp) { /* * If prev_bp is set, this is an overflow page, hook it in to * the buffer overflow links. */ #ifdef DEBUG1 (void)fprintf(stderr, "NEWBUF2: %d->ovfl was %d is now %d\n", prev_bp->addr, (prev_bp->ovfl ? prev_bp->ovfl->addr : 0), (bp ? bp->addr : 0)); #endif prev_bp->ovfl = bp; bp->flags = 0; } else bp->flags = BUF_BUCKET; MRU_INSERT(bp); return (bp); } void __buf_init(HTAB *hashp, int nbytes) { BUFHEAD *bfp; int npages; bfp = &(hashp->bufhead); npages = (nbytes + hashp->BSIZE - 1) >> hashp->BSHIFT; npages = MAX(npages, MIN_BUFFERS); hashp->nbufs = npages; bfp->next = bfp; bfp->prev = bfp; /* * This space is calloc'd so these are already null. * * bfp->ovfl = NULL; * bfp->flags = 0; * bfp->page = NULL; * bfp->addr = 0; */ } int __buf_free(HTAB *hashp, int do_free, int to_disk) { BUFHEAD *bp; /* Need to make sure that buffer manager has been initialized */ if (!LRU) return (0); for (bp = LRU; bp != &hashp->bufhead;) { /* Check that the buffer is valid */ if (bp->addr || IS_BUCKET(bp->flags)) { if (to_disk && (bp->flags & BUF_MOD) && __put_page(hashp, bp->page, bp->addr, IS_BUCKET(bp->flags), 0)) return (-1); } /* Check if we are freeing stuff */ if (do_free) { if (bp->page) { (void)memset(bp->page, 0, hashp->BSIZE); free(bp->page); } BUF_REMOVE(bp); free(bp); bp = LRU; } else bp = bp->prev; } return (0); } void __reclaim_buf(HTAB *hashp, BUFHEAD *bp) { bp->ovfl = 0; bp->addr = 0; bp->flags = 0; BUF_REMOVE(bp); LRU_INSERT(bp); }