store_dir_aufs.c

一个功能非常全面的代理服务器源代码程序,

/*
 * $Id: store_dir_aufs.c,v 1.64 2006/09/15 20:13:02 serassio Exp $
 *
 * DEBUG: section 47    Store Directory Routines
 * AUTHOR: Duane Wessels
 *
 * SQUID Web Proxy Cache          http://www.squid-cache.org/
 * ----------------------------------------------------------
 *
 *  Squid is the result of efforts by numerous individuals from
 *  the Internet community; see the CONTRIBUTORS file for full
 *  details.   Many organizations have provided support for Squid's
 *  development; see the SPONSORS file for full details.  Squid is
 *  Copyrighted (C) 2001 by the Regents of the University of
 *  California; see the COPYRIGHT file for full details.  Squid
 *  incorporates software developed and/or copyrighted by other
 *  sources; 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"

#include "async_io.h"
#include "store_asyncufs.h"

#define DefaultLevelOneDirs     16
#define DefaultLevelTwoDirs     256
#define STORE_META_BUFSZ 4096

typedef struct _RebuildState RebuildState;
struct _RebuildState {
    SwapDir *sd;
    int n_read;
    FILE *log;
    int speed;
    int curlvl1;
    int curlvl2;
    struct {
	unsigned int need_to_validate:1;
	unsigned int clean:1;
	unsigned int init:1;
    } flags;
    int done;
    int in_dir;
    int fn;
    struct dirent *entry;
    DIR *td;
    char fullpath[SQUID_MAXPATHLEN];
    char fullfilename[SQUID_MAXPATHLEN];
    struct _store_rebuild_data counts;
};

int n_asyncufs_dirs = 0;
static int *asyncufs_dir_index = NULL;
MemPool *squidaio_state_pool = NULL;
MemPool *aufs_qread_pool = NULL;
MemPool *aufs_qwrite_pool = NULL;
static int asyncufs_initialised = 0;

static char *storeAufsDirSwapSubDir(SwapDir *, int subdirn);
static int storeAufsDirCreateDirectory(const char *path, int);
static int storeAufsDirVerifyCacheDirs(SwapDir *);
static int storeAufsDirVerifyDirectory(const char *path);
static void storeAufsDirCreateSwapSubDirs(SwapDir *);
static char *storeAufsDirSwapLogFile(SwapDir *, const char *);
static EVH storeAufsDirRebuildFromDirectory;
static EVH storeAufsDirRebuildFromSwapLog;
static int storeAufsDirGetNextFile(RebuildState *, sfileno *, int *size);
static StoreEntry *storeAufsDirAddDiskRestore(SwapDir * SD, const cache_key * key,
    sfileno file_number,
    squid_file_sz swap_file_sz,
    time_t expires,
    time_t timestamp,
    time_t lastref,
    time_t lastmod,
    u_num32 refcount,
    u_short flags,
    int clean);
static void storeAufsDirRebuild(SwapDir * sd);
static void storeAufsDirCloseTmpSwapLog(SwapDir * sd);
static FILE *storeAufsDirOpenTmpSwapLog(SwapDir *, int *, int *);
static STLOGOPEN storeAufsDirOpenSwapLog;
static STINIT storeAufsDirInit;
static STFREE storeAufsDirFree;
static STLOGCLEANSTART storeAufsDirWriteCleanStart;
static STLOGCLEANNEXTENTRY storeAufsDirCleanLogNextEntry;
static STLOGCLEANWRITE storeAufsDirWriteCleanEntry;
static STLOGCLEANDONE storeAufsDirWriteCleanDone;
static STLOGCLOSE storeAufsDirCloseSwapLog;
static STLOGWRITE storeAufsDirSwapLog;
static STNEWFS storeAufsDirNewfs;
static STDUMP storeAufsDirDump;
static STMAINTAINFS storeAufsDirMaintain;
static STCHECKOBJ storeAufsDirCheckObj;
static STCHECKLOADAV storeAufsDirCheckLoadAv;
static STREFOBJ storeAufsDirRefObj;
static STUNREFOBJ storeAufsDirUnrefObj;
static QS rev_int_sort;
static int storeAufsDirClean(int swap_index);
static EVH storeAufsDirCleanEvent;
static int storeAufsFilenoBelongsHere(int fn, int F0, int F1, int F2);
static int storeAufsCleanupDoubleCheck(SwapDir *, StoreEntry *);
static void storeAufsDirStats(SwapDir *, StoreEntry *);
static void storeAufsDirInitBitmap(SwapDir *);
static int storeAufsDirValidFileno(SwapDir *, sfileno, int);

/* The MAIN externally visible function */
STSETUP storeFsSetup_aufs;

/*
 * These functions were ripped straight out of the heart of store_dir.c.
 * They assume that the given filenum is on a asyncufs partiton, which may or
 * may not be true.. 
 * XXX this evilness should be tidied up at a later date!
 */

static int
storeAufsDirMapBitTest(SwapDir * SD, sfileno filn)
{
    squidaioinfo_t *aioinfo;
    aioinfo = (squidaioinfo_t *) SD->fsdata;
    return file_map_bit_test(aioinfo->map, filn);
}

static void
storeAufsDirMapBitSet(SwapDir * SD, sfileno filn)
{
    squidaioinfo_t *aioinfo;
    aioinfo = (squidaioinfo_t *) SD->fsdata;
    file_map_bit_set(aioinfo->map, filn);
}

void
storeAufsDirMapBitReset(SwapDir * SD, sfileno filn)
{
    squidaioinfo_t *aioinfo;
    aioinfo = (squidaioinfo_t *) SD->fsdata;
    /*
     * We have to test the bit before calling file_map_bit_reset.
     * file_map_bit_reset doesn't do bounds checking.  It assumes
     * filn is a valid file number, but it might not be because
     * the map is dynamic in size.  Also clearing an already clear
     * bit puts the map counter of-of-whack.
     */
    if (file_map_bit_test(aioinfo->map, filn))
	file_map_bit_reset(aioinfo->map, filn);
}

int
storeAufsDirMapBitAllocate(SwapDir * SD)
{
    squidaioinfo_t *aioinfo = (squidaioinfo_t *) SD->fsdata;
    int fn;
    fn = file_map_allocate(aioinfo->map, aioinfo->suggest);
    file_map_bit_set(aioinfo->map, fn);
    aioinfo->suggest = fn + 1;
    return fn;
}

/*
 * Initialise the asyncufs bitmap
 *
 * If there already is a bitmap, and the numobjects is larger than currently
 * configured, we allocate a new bitmap and 'grow' the old one into it.
 */
static void
storeAufsDirInitBitmap(SwapDir * sd)
{
    squidaioinfo_t *aioinfo = (squidaioinfo_t *) sd->fsdata;

    if (aioinfo->map == NULL) {
	/* First time */
	aioinfo->map = file_map_create();
    } else if (aioinfo->map->max_n_files) {
	/* it grew, need to expand */
	/* XXX We don't need it anymore .. */
    }
    /* else it shrunk, and we leave the old one in place */
}

static char *
storeAufsDirSwapSubDir(SwapDir * sd, int subdirn)
{
    squidaioinfo_t *aioinfo = (squidaioinfo_t *) sd->fsdata;

    LOCAL_ARRAY(char, fullfilename, SQUID_MAXPATHLEN);
    assert(0 <= subdirn && subdirn < aioinfo->l1);
    snprintf(fullfilename, SQUID_MAXPATHLEN, "%s/%02X", sd->path, subdirn);
    return fullfilename;
}

static int
storeAufsDirCreateDirectory(const char *path, int should_exist)
{
    int created = 0;
    struct stat st;
    getCurrentTime();
    if (0 == stat(path, &st)) {
	if (S_ISDIR(st.st_mode)) {
	    debug(47, should_exist ? 3 : 1) ("%s exists\n", path);
	} else {
	    fatalf("Swap directory %s is not a directory.", path);
	}
#ifdef _SQUID_MSWIN_
    } else if (0 == mkdir(path)) {
#else
    } else if (0 == mkdir(path, 0755)) {
#endif
	debug(47, should_exist ? 1 : 3) ("%s created\n", path);
	created = 1;
    } else {
	fatalf("Failed to make swap directory %s: %s",
	    path, xstrerror());
    }
    return created;
}

static int
storeAufsDirVerifyDirectory(const char *path)
{
    struct stat sb;
    if (stat(path, &sb) < 0) {
	debug(47, 0) ("%s: %s\n", path, xstrerror());
	return -1;
    }
    if (S_ISDIR(sb.st_mode) == 0) {
	debug(47, 0) ("%s is not a directory\n", path);
	return -1;
    }
    return 0;
}

/*
 * This function is called by storeAufsDirInit().  If this returns < 0,
 * then Squid exits, complains about swap directories not
 * existing, and instructs the admin to run 'squid -z'
 */
static int
storeAufsDirVerifyCacheDirs(SwapDir * sd)
{
    squidaioinfo_t *aioinfo = (squidaioinfo_t *) sd->fsdata;
    int j;
    const char *path = sd->path;

    if (storeAufsDirVerifyDirectory(path) < 0)
	return -1;
    for (j = 0; j < aioinfo->l1; j++) {
	path = storeAufsDirSwapSubDir(sd, j);
	if (storeAufsDirVerifyDirectory(path) < 0)
	    return -1;
    }
    return 0;
}

static void
storeAufsDirCreateSwapSubDirs(SwapDir * sd)
{
    squidaioinfo_t *aioinfo = (squidaioinfo_t *) sd->fsdata;
    int i, k;
    int should_exist;
    LOCAL_ARRAY(char, name, MAXPATHLEN);
    for (i = 0; i < aioinfo->l1; i++) {
	snprintf(name, MAXPATHLEN, "%s/%02X", sd->path, i);
	if (storeAufsDirCreateDirectory(name, 0))
	    should_exist = 0;
	else
	    should_exist = 1;
	debug(47, 1) ("Making directories in %s\n", name);
	for (k = 0; k < aioinfo->l2; k++) {
	    snprintf(name, MAXPATHLEN, "%s/%02X/%02X", sd->path, i, k);
	    storeAufsDirCreateDirectory(name, should_exist);
	}
    }
}

static char *
storeAufsDirSwapLogFile(SwapDir * sd, const char *ext)
{
    LOCAL_ARRAY(char, path, SQUID_MAXPATHLEN);
    LOCAL_ARRAY(char, pathtmp, SQUID_MAXPATHLEN);
    LOCAL_ARRAY(char, digit, 32);
    char *pathtmp2;
    if (Config.Log.swap) {
	xstrncpy(pathtmp, sd->path, SQUID_MAXPATHLEN - 64);
	pathtmp2 = pathtmp;
	while ((pathtmp2 = strchr(pathtmp2, '/')) != NULL)
	    *pathtmp2 = '.';
	while (strlen(pathtmp) && pathtmp[strlen(pathtmp) - 1] == '.')
	    pathtmp[strlen(pathtmp) - 1] = '\0';
	for (pathtmp2 = pathtmp; *pathtmp2 == '.'; pathtmp2++);
	snprintf(path, SQUID_MAXPATHLEN - 64, Config.Log.swap, pathtmp2);
	if (strncmp(path, Config.Log.swap, SQUID_MAXPATHLEN - 64) == 0) {
	    strcat(path, ".");
	    snprintf(digit, 32, "%02d", sd->index);
	    strncat(path, digit, 3);
	}
    } else {
	xstrncpy(path, sd->path, SQUID_MAXPATHLEN - 64);
	strcat(path, "/swap.state");
    }
    if (ext)
	strncat(path, ext, 16);
    return path;
}

static void
storeAufsDirOpenSwapLog(SwapDir * sd)
{
    squidaioinfo_t *aioinfo = (squidaioinfo_t *) sd->fsdata;
    char *path;
    int fd;
    path = storeAufsDirSwapLogFile(sd, NULL);
    if (aioinfo->swaplog_fd >= 0) {
	debug(50, 1) ("storeAufsDirOpenSwapLog: %s already open\n", path);
	return;
    }
    fd = file_open(path, O_WRONLY | O_CREAT | O_BINARY);
    if (fd < 0) {
	debug(50, 1) ("%s: %s\n", path, xstrerror());
	fatal("storeAufsDirOpenSwapLog: Failed to open swap log.");
    }
    debug(50, 3) ("Cache Dir #%d log opened on FD %d\n", sd->index, fd);
    aioinfo->swaplog_fd = fd;
}

static void
storeAufsDirCloseSwapLog(SwapDir * sd)
{
    squidaioinfo_t *aioinfo = (squidaioinfo_t *) sd->fsdata;
    if (aioinfo->swaplog_fd < 0)	/* not open */
	return;
    file_close(aioinfo->swaplog_fd);
    debug(47, 3) ("Cache Dir #%d log closed on FD %d\n",
	sd->index, aioinfo->swaplog_fd);
    aioinfo->swaplog_fd = -1;
}

static void
storeAufsCheckConfig(SwapDir * sd)
{
    if (!opt_create_swap_dirs)
	requirePathnameExists("cache_dir", sd->path);
}


static void
storeAufsDirInit(SwapDir * sd)
{
    static int started_clean_event = 0;
    static const char *errmsg =
    "\tFailed to verify one of the swap directories, Check cache.log\n"
    "\tfor details.  Run 'squid -z' to create swap directories\n"
    "\tif needed, or if running Squid for the first time.";
    storeAufsDirInitBitmap(sd);
    if (storeAufsDirVerifyCacheDirs(sd) < 0)
	fatal(errmsg);
    squidaio_init();
    storeAufsDirOpenSwapLog(sd);
    storeAufsDirRebuild(sd);
    if (!started_clean_event) {
	eventAdd("storeDirClean", storeAufsDirCleanEvent, NULL, 15.0, 1);
	started_clean_event = 1;
    }
    (void) storeDirGetBlkSize(sd->path, &sd->fs.blksize);
}

static void
storeAufsDirRebuildComplete(RebuildState * rb)
{
    if (rb->log) {
	debug(47, 1) ("Done reading %s swaplog (%d entries)\n",
	    rb->sd->path, rb->n_read);
	fclose(rb->log);
	rb->log = NULL;
    } else {
	debug(47, 1) ("Done scanning %s (%d entries)\n",
	    rb->sd->path, rb->counts.scancount);
    }
    store_dirs_rebuilding--;
    storeAufsDirCloseTmpSwapLog(rb->sd);
    storeRebuildComplete(&rb->counts);
    cbdataFree(rb);
}

static void
storeAufsDirRebuildFromDirectory(void *data)
{
    RebuildState *rb = data;
    SwapDir *SD = rb->sd;
    LOCAL_ARRAY(char, hdr_buf, SM_PAGE_SIZE);
    StoreEntry *e = NULL;
    StoreEntry tmpe;
    cache_key key[MD5_DIGEST_CHARS];
    sfileno filn = 0;
    int count;
    int size;
    struct stat sb;
    int swap_hdr_len;
    int fd = -1;
    tlv *tlv_list;
    tlv *t;
    assert(rb != NULL);
    debug(47, 3) ("storeAufsDirRebuildFromDirectory: DIR #%d\n", rb->sd->index);
    for (count = 0; count < rb->speed; count++) {
	assert(fd == -1);
	fd = storeAufsDirGetNextFile(rb, &filn, &size);
	if (fd == -2) {
	    storeAufsDirRebuildComplete(rb);
	    return;
	} else if (fd < 0) {
	    continue;
	}
	assert(fd > -1);
	/* lets get file stats here */
	if (fstat(fd, &sb) < 0) {
	    debug(47, 1) ("storeAufsDirRebuildFromDirectory: fstat(FD %d): %s\n",
		fd, xstrerror());
	    file_close(fd);
	    store_open_disk_fd--;
	    fd = -1;
	    continue;
	}
	if ((++rb->counts.scancount & 0xFFFF) == 0)
	    debug(47, 3) ("  %s %7d files opened so far.\n",
		rb->sd->path, rb->counts.scancount);
	debug(47, 9) ("file_in: fd=%d %08X\n", fd, filn);
	statCounter.syscalls.disk.reads++;
	if (FD_READ_METHOD(fd, hdr_buf, SM_PAGE_SIZE) < 0) {
	    debug(47, 1) ("storeAufsDirRebuildFromDirectory: 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
	if (sb.st_size == 0)
	    continue;
#endif
	tlv_list = storeSwapMetaUnpack(hdr_buf, &swap_hdr_len);
	if (tlv_list == NULL) {
	    debug(47, 1) ("storeAufsDirRebuildFromDirectory: failed to get meta data\n");
	    /* XXX shouldn't this be a call to storeAufsUnlink ? */
	    storeAufsDirUnlinkFile(SD, filn);
	    continue;
	}
	debug(47, 3) ("storeAufsDirRebuildFromDirectory: 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;
#if SIZEOF_SQUID_FILE_SZ == SIZEOF_SIZE_T
	    case STORE_META_STD:
		assert(t->length == STORE_HDR_METASIZE);
		xmemcpy(&tmpe.timestamp, t->value, STORE_HDR_METASIZE);
		break;
#else
	    case STORE_META_STD_LFS:
		assert(t->length == STORE_HDR_METASIZE);
		xmemcpy(&tmpe.timestamp, t->value, STORE_HDR_METASIZE);
		break;
	    case STORE_META_STD:
		assert(t->length == STORE_HDR_METASIZE_OLD);
		{
		    struct {
			time_t timestamp;