store_rebuild.c

-

/*
 * $Id: store_rebuild.c,v 1.56.2.1 1999/04/27 21:54:24 wessels Exp $
 *
 * DEBUG: section 20    Store Rebuild Routines
 * AUTHOR: Duane Wessels
 *
 * SQUID Internet Object Cache  http://squid.nlanr.net/Squid/
 * ----------------------------------------------------------
 *
 *  Squid is the result of efforts by numerous individuals from the
 *  Internet community.  Development is led by Duane Wessels of the
 *  National Laboratory for Applied Network Research and funded by the
 *  National Science Foundation.  Squid is Copyrighted (C) 1998 by
 *  Duane Wessels and the University of California San Diego.  Please
 *  see the COPYRIGHT file for full details.  Squid incorporates
 *  software developed and/or copyrighted by other sources.  Please see
 *  the CREDITS file for full details.
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; either version 2 of the License, or
 *  (at your option) any later version.
 *  
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *  
 *  You should have received a copy of the GNU General Public License
 *  along with this program; if not, write to the Free Software
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111, USA.
 *
 */

#include "squid.h"

#define STORE_META_BUFSZ 4096


typedef struct _rebuild_dir rebuild_dir;
typedef int RBHD(rebuild_dir * d);

struct _rebuild_dir {
    int dirn;
    int n_read;
    FILE *log;
    int speed;
    int clean;
    int curlvl1;
    int curlvl2;
    int flag;
    int done;
    int in_dir;
    int fn;
    struct dirent *entry;
    DIR *td;
    RBHD *rebuild_func;
    rebuild_dir *next;
    char fullpath[SQUID_MAXPATHLEN];
    char fullfilename[SQUID_MAXPATHLEN];
};

struct storeRebuildState {
    rebuild_dir *rebuild_dir;
    int objcount;		/* # objects successfully reloaded */
    int expcount;		/* # objects expired */
    int linecount;		/* # lines parsed from cache logfile */
    int statcount;		/* # entries from directory walking */
    int clashcount;		/* # swapfile clashes avoided */
    int dupcount;		/* # duplicates purged */
    int cancelcount;		/* # SWAP_LOG_DEL objects purged */
    int invalid;		/* # bad lines */
    int badflags;		/* # bad e->flags */
    int need_to_validate;
    int bad_log_op;
    int zero_object_sz;
    time_t start;
    time_t stop;
} RebuildState;

typedef struct valid_ctrl_t {
    struct stat *sb;
    StoreEntry *e;
    STVLDCB *callback;
    void *callback_data;
} valid_ctrl_t;

static RBHD storeRebuildFromDirectory;
static RBHD storeRebuildFromSwapLog;
static void storeRebuildComplete(void);
static EVH storeRebuildADirectory;
static int storeGetNextFile(rebuild_dir *, int *sfileno, int *size);
static StoreEntry *storeAddDiskRestore(const cache_key * key,
    int file_number,
    size_t swap_file_sz,
    time_t expires,
    time_t timestamp,
    time_t lastref,
    time_t lastmod,
    u_num32 refcount,
    u_short flags,
    int clean);
static AIOCB storeValidateComplete;

static int
storeRebuildFromDirectory(rebuild_dir * d)
{
    LOCAL_ARRAY(char, hdr_buf, DISK_PAGE_SIZE);
    StoreEntry *e = NULL;
    StoreEntry tmpe;
    cache_key key[MD5_DIGEST_CHARS];
    int sfileno = 0;
    int count;
    int size;
    struct stat sb;
    int swap_hdr_len;
    int fd = -1;
    tlv *tlv_list;
    tlv *t;
    assert(d != NULL);
    debug(20, 3) ("storeRebuildFromDirectory: DIR #%d\n", d->dirn);
    for (count = 0; count < d->speed; count++) {
	assert(fd == -1);
	fd = storeGetNextFile(d, &sfileno, &size);
	if (fd == -2) {
	    debug(20, 1) ("storeRebuildFromDirectory: DIR #%d done!\n", d->dirn);
	    storeDirCloseTmpSwapLog(d->dirn);
	    store_rebuilding = 0;
	    return -1;
	} else if (fd < 0) {
	    continue;
	}
	assert(fd > -1);
	/* lets get file stats here */
	if (fstat(fd, &sb) < 0) {
	    debug(20, 1) ("storeRebuildFromDirectory: fstat(FD %d): %s\n",
		fd, xstrerror());
	    file_close(fd);
	    store_open_disk_fd--;
	    fd = -1;
	    continue;
	}
	if ((++RebuildState.statcount & 0xFFFF) == 0)
	    debug(20, 1) ("  %7d files opened so far.\n",
		RebuildState.statcount);
	debug(20, 9) ("file_in: fd=%d %08X\n", fd, sfileno);
	Counter.syscalls.disk.reads++;
	if (read(fd, hdr_buf, DISK_PAGE_SIZE) < 0) {
	    debug(20, 1) ("storeRebuildFromDirectory: read(FD %d): %s\n",
		fd, xstrerror());
	    file_close(fd);
	    store_open_disk_fd--;
	    fd = -1;
	    continue;
	}
	file_close(fd);
	store_open_disk_fd--;
	fd = -1;
	swap_hdr_len = 0;
#if USE_TRUNCATE_NOT_UNLINK
	if (sb.st_size == 0)
	    continue;
#endif
	tlv_list = storeSwapMetaUnpack(hdr_buf, &swap_hdr_len);
	if (tlv_list == NULL) {
	    debug(20, 1) ("storeRebuildFromDirectory: failed to get meta data\n");
	    storeUnlinkFileno(sfileno);
	    continue;
	}
	debug(20, 3) ("storeRebuildFromDirectory: successful swap meta unpacking\n");
	memset(key, '\0', MD5_DIGEST_CHARS);
	memset(&tmpe, '\0', sizeof(StoreEntry));
	for (t = tlv_list; t; t = t->next) {
	    switch (t->type) {
	    case STORE_META_KEY:
		assert(t->length == MD5_DIGEST_CHARS);
		xmemcpy(key, t->value, MD5_DIGEST_CHARS);
		break;
	    case STORE_META_STD:
		assert(t->length == STORE_HDR_METASIZE);
		xmemcpy(&tmpe.timestamp, t->value, STORE_HDR_METASIZE);
		break;
	    default:
		break;
	    }
	}
	storeSwapTLVFree(tlv_list);
	tlv_list = NULL;
	if (storeKeyNull(key)) {
	    debug(20, 1) ("storeRebuildFromDirectory: NULL key\n");
	    storeUnlinkFileno(sfileno);
	    continue;
	}
	tmpe.key = key;
	/* check sizes */
	if (tmpe.swap_file_sz == 0) {
	    tmpe.swap_file_sz = sb.st_size;
	} else if (tmpe.swap_file_sz == sb.st_size - swap_hdr_len) {
	    tmpe.swap_file_sz = sb.st_size;
	} else if (tmpe.swap_file_sz != sb.st_size) {
	    debug(20, 1) ("storeRebuildFromDirectory: SIZE MISMATCH %d!=%d\n",
		tmpe.swap_file_sz, (int) sb.st_size);
	    storeUnlinkFileno(sfileno);
	    continue;
	}
	if (EBIT_TEST(tmpe.flags, KEY_PRIVATE)) {
	    storeUnlinkFileno(sfileno);
	    RebuildState.badflags++;
	    continue;
	}
	e = storeGet(key);
	if (e && e->lastref >= tmpe.lastref) {
	    /* key already exists, current entry is newer */
	    /* keep old, ignore new */
	    RebuildState.dupcount++;
	    continue;
	} else if (NULL != e) {
	    /* URL already exists, this swapfile not being used */
	    /* junk old, load new */
	    storeRelease(e);	/* release old entry */
	    RebuildState.dupcount++;
	}
	RebuildState.objcount++;
	storeEntryDump(&tmpe, 5);
	e = storeAddDiskRestore(key,
	    sfileno,
	    tmpe.swap_file_sz,
	    tmpe.expires,
	    tmpe.timestamp,
	    tmpe.lastref,
	    tmpe.lastmod,
	    tmpe.refcount,	/* refcount */
	    tmpe.flags,		/* flags */
	    d->clean);
    }
    return count;
}

static int
storeRebuildFromSwapLog(rebuild_dir * d)
{
    StoreEntry *e = NULL;
    storeSwapLogData s;
    size_t ss = sizeof(storeSwapLogData);
    int count;
    int used;			/* is swapfile already in use? */
    int disk_entry_newer;	/* is the log entry newer than current entry? */
    double x;
    assert(d != NULL);
    /* load a number of objects per invocation */
    for (count = 0; count < d->speed; count++) {
	if (fread(&s, ss, 1, d->log) != 1) {
	    debug(20, 1) ("Done reading Cache Dir #%d swaplog (%d entries)\n",
		d->dirn, d->n_read);
	    fclose(d->log);
	    d->log = NULL;
	    storeDirCloseTmpSwapLog(d->dirn);
	    return -1;
	}
	d->n_read++;
	if (s.op <= SWAP_LOG_NOP)
	    continue;
	if (s.op >= SWAP_LOG_MAX)
	    continue;
	s.swap_file_number = storeDirProperFileno(d->dirn, s.swap_file_number);
	debug(20, 3) ("storeRebuildFromSwapLog: %s %s %08X\n",
	    swap_log_op_str[(int) s.op],
	    storeKeyText(s.key),
	    s.swap_file_number);
	if (s.op == SWAP_LOG_ADD) {
	    (void) 0;
	} else if (s.op == SWAP_LOG_DEL) {
	    if ((e = storeGet(s.key)) != NULL) {
		/*
		 * Make sure we don't unlink the file, it might be
		 * in use by a subsequent entry.  Also note that
		 * we don't have to subtract from store_swap_size
		 * because adding to store_swap_size happens in
		 * the cleanup procedure.
		 */
		storeExpireNow(e);
		storeReleaseRequest(e);
		if (e->swap_file_number > -1) {
		    storeDirMapBitReset(e->swap_file_number);
		    e->swap_file_number = -1;
		}
		RebuildState.objcount--;
		RebuildState.cancelcount++;
	    }
	    continue;
	} else {
	    x = log(++RebuildState.bad_log_op) / log(10.0);
	    if (0.0 == x - (double) (int) x)
		debug(20, 1) ("WARNING: %d invalid swap log entries found\n",
		    RebuildState.bad_log_op);
	    RebuildState.invalid++;
	    continue;
	}
	if ((++RebuildState.linecount & 0xFFFF) == 0)
	    debug(20, 1) ("  %7d Entries read so far.\n",
		RebuildState.linecount);
	if (!storeDirValidFileno(s.swap_file_number)) {
	    RebuildState.invalid++;
	    continue;
	}
	if (EBIT_TEST(s.flags, KEY_PRIVATE)) {
	    RebuildState.badflags++;
	    continue;
	}
	e = storeGet(s.key);
	used = storeDirMapBitTest(s.swap_file_number);
	/* If this URL already exists in the cache, does the swap log
	 * appear to have a newer entry?  Compare 'lastref' from the
	 * swap log to e->lastref. */
	disk_entry_newer = e ? (s.lastref > e->lastref ? 1 : 0) : 0;
	if (used && !disk_entry_newer) {
	    /* log entry is old, ignore it */
	    RebuildState.clashcount++;
	    continue;
	} else if (used && e && e->swap_file_number == s.swap_file_number) {
	    /* swapfile taken, same URL, newer, update meta */
	    if (e->store_status == STORE_OK) {
		e->lastref = s.timestamp;
		e->timestamp = s.timestamp;
		e->expires = s.expires;
		e->lastmod = s.lastmod;
		e->flags = s.flags;
		e->refcount += s.refcount;
	    } else {
		debug_trap("storeRebuildFromSwapLog: bad condition");
		debug(20, 1) ("\tSee %s:%d\n", __FILE__, __LINE__);
	    }
	    continue;
	} else if (used) {
	    /* swapfile in use, not by this URL, log entry is newer */
	    /* This is sorta bad: the log entry should NOT be newer at this
	     * point.  If the log is dirty, the filesize check should have
	     * caught this.  If the log is clean, there should never be a
	     * newer entry. */
	    debug(20, 1) ("WARNING: newer swaplog entry for fileno %08X\n",
		s.swap_file_number);
	    /* I'm tempted to remove the swapfile here just to be safe,
	     * but there is a bad race condition in the NOVM version if
	     * the swapfile has recently been opened for writing, but
	     * not yet opened for reading.  Because we can't map
	     * swapfiles back to StoreEntrys, we don't know the state
	     * of the entry using that file.  */
	    /* We'll assume the existing entry is valid, probably because
	     * were in a slow rebuild and the the swap file number got taken
	     * and the validation procedure hasn't run. */
	    assert(RebuildState.need_to_validate);
	    RebuildState.clashcount++;
	    continue;
	} else if (e && !disk_entry_newer) {
	    /* key already exists, current entry is newer */
	    /* keep old, ignore new */
	    RebuildState.dupcount++;
	    continue;
	} else if (e) {
	    /* key already exists, this swapfile not being used */
	    /* junk old, load new */
	    storeExpireNow(e);
	    storeReleaseRequest(e);
	    if (e->swap_file_number > -1) {
		storeDirMapBitReset(e->swap_file_number);
		e->swap_file_number = -1;
	    }
	    RebuildState.dupcount++;
	} else {
	    /* URL doesnt exist, swapfile not in use */
	    /* load new */
	    (void) 0;
	}
	/* update store_swap_size */
	RebuildState.objcount++;
	e = storeAddDiskRestore(s.key,
	    s.swap_file_number,
	    s.swap_file_sz,
	    s.expires,
	    s.timestamp,
	    s.lastref,
	    s.lastmod,
	    s.refcount,
	    s.flags,
	    d->clean);
	storeDirSwapLog(e, SWAP_LOG_ADD);
    }
    return count;
}