mp_fput.c

/*-
 * See the file LICENSE for redistribution information.
 *
 * Copyright (c) 1996-2005
 *      Sleepycat Software.  All rights reserved.
 *
 * $Id: mp_fput.c,v 12.7 2005/10/07 20:21:33 ubell Exp $
 */

#include "db_config.h"

#ifndef NO_SYSTEM_INCLUDES
#include <sys/types.h>

#endif

#include "db_int.h"
#include "dbinc/db_shash.h"
#include "dbinc/log.h"
#include "dbinc/mp.h"

static int __memp_reset_lru __P((DB_ENV *, REGINFO *));

/*
 * __memp_fput_pp --
 *      DB_MPOOLFILE->put pre/post processing.
 *
 * PUBLIC: int __memp_fput_pp __P((DB_MPOOLFILE *, void *, u_int32_t));
 */
int
__memp_fput_pp(dbmfp, pgaddr, flags)
        DB_MPOOLFILE *dbmfp;
        void *pgaddr;
        u_int32_t flags;
{
        DB_ENV *dbenv;
        DB_THREAD_INFO *ip;
        int ret, t_ret;

        dbenv = dbmfp->dbenv;
        PANIC_CHECK(dbenv);

        ENV_ENTER(dbenv, ip);

        ret = __memp_fput(dbmfp, pgaddr, flags);
        if (IS_ENV_REPLICATED(dbenv) &&
            (t_ret = __op_rep_exit(dbenv)) != 0 && ret == 0)
                ret = t_ret;

        ENV_LEAVE(dbenv, ip);
        return (ret);
}

/*
 * __memp_fput --
 *      DB_MPOOLFILE->put.
 *
 * PUBLIC: int __memp_fput __P((DB_MPOOLFILE *, void *, u_int32_t));
 */
int
__memp_fput(dbmfp, pgaddr, flags)
        DB_MPOOLFILE *dbmfp;
        void *pgaddr;
        u_int32_t flags;
{
        BH *fbhp, *bhp, *prev;
        DB_ENV *dbenv;
        DB_MPOOL *dbmp;
        DB_MPOOL_HASH *hp;
        MPOOL *c_mp;
        MPOOLFILE *mfp;
        u_int32_t n_cache;
        int adjust, ret, t_ret;

        dbenv = dbmfp->dbenv;
        MPF_ILLEGAL_BEFORE_OPEN(dbmfp, "DB_MPOOLFILE->put");
        dbmp = dbenv->mp_handle;
        ret = 0;

        /*
         * Check arguments, but don't fail because we want to unpin the page
         * regardless.  The problem is when running with replication.  There
         * is a reference count we incremented when __memp_fget was called,
         * and we need to unpin the page and decrement that reference count.
         * If we see flag problems, mark the page dirty.
         */
        if (flags) {
                if (__db_fchk(dbenv, "memp_fput", flags,
                    DB_MPOOL_CLEAN | DB_MPOOL_DIRTY | DB_MPOOL_DISCARD) != 0 ||
                    __db_fcchk(dbenv, "memp_fput", flags,
                    DB_MPOOL_CLEAN, DB_MPOOL_DIRTY) != 0) {
                        flags = DB_MPOOL_DIRTY;
                        ret = EINVAL;
                        DB_ASSERT(0);
                }

                if (LF_ISSET(DB_MPOOL_DIRTY) && F_ISSET(dbmfp, MP_READONLY)) {
                        __db_err(dbenv,
                            "%s: dirty flag set for readonly file page",
                            __memp_fn(dbmfp));
                        flags = 0;
                        ret = EINVAL;
                        DB_ASSERT(0);
                }
        }

        /*
         * If we're mapping the file, there's nothing to do.  Because we can
         * stop mapping the file at any time, we have to check on each buffer
         * to see if the address we gave the application was part of the map
         * region.
         */
        if (dbmfp->addr != NULL && pgaddr >= dbmfp->addr &&
            (u_int8_t *)pgaddr <= (u_int8_t *)dbmfp->addr + dbmfp->len)
                return (0);

#ifdef DIAGNOSTIC
        /*
         * Decrement the per-file pinned buffer count (mapped pages aren't
         * counted).
         */
        MPOOL_SYSTEM_LOCK(dbenv);
        if (dbmfp->pinref == 0) {
                MPOOL_SYSTEM_UNLOCK(dbenv);
                __db_err(dbenv,
                    "%s: more pages returned than retrieved", __memp_fn(dbmfp));
                return (__db_panic(dbenv, EACCES));
        }
        --dbmfp->pinref;
        MPOOL_SYSTEM_UNLOCK(dbenv);
#endif

        /* Convert a page address to a buffer header and hash bucket. */
        bhp = (BH *)((u_int8_t *)pgaddr - SSZA(BH, buf));
        n_cache = NCACHE(dbmp->reginfo[0].primary, bhp->mf_offset, bhp->pgno);
        c_mp = dbmp->reginfo[n_cache].primary;
        hp = R_ADDR(&dbmp->reginfo[n_cache], c_mp->htab);
        hp = &hp[NBUCKET(c_mp, bhp->mf_offset, bhp->pgno)];

        MUTEX_LOCK(dbenv, hp->mtx_hash);

        /* Set/clear the page bits. */
        if (LF_ISSET(DB_MPOOL_CLEAN) &&
            F_ISSET(bhp, BH_DIRTY) && !F_ISSET(bhp, BH_DIRTY_CREATE)) {
                DB_ASSERT(hp->hash_page_dirty != 0);
                --hp->hash_page_dirty;
                F_CLR(bhp, BH_DIRTY);
        }
        if (LF_ISSET(DB_MPOOL_DIRTY) && !F_ISSET(bhp, BH_DIRTY)) {
                ++hp->hash_page_dirty;
                F_SET(bhp, BH_DIRTY);
        }
        if (LF_ISSET(DB_MPOOL_DISCARD))
                F_SET(bhp, BH_DISCARD);

        /*
         * Check for a reference count going to zero.  This can happen if the
         * application returns a page twice.
         */
        if (bhp->ref == 0) {
                MUTEX_UNLOCK(dbenv, hp->mtx_hash);
                __db_err(dbenv, "%s: page %lu: unpinned page returned",
                    __memp_fn(dbmfp), (u_long)bhp->pgno);
                return (__db_panic(dbenv, EACCES));
        }

        /* Note the activity so allocation won't decide to quit. */
        ++c_mp->put_counter;

        /*
         * Mark the file dirty.  Check for a dirty bit on the buffer as well
         * as the dirty flag because the buffer might have been marked dirty
         * in the DB_MPOOLFILE->set method.
         */
        mfp = dbmfp->mfp;
        if (LF_ISSET(DB_MPOOL_DIRTY) || F_ISSET(bhp, BH_DIRTY))
                mfp->file_written = 1;

        /*
         * If more than one reference to the page or a reference other than a
         * thread waiting to flush the buffer to disk, we're done.  Ignore the
         * discard flags (for now) and leave the buffer's priority alone.
         */
        if (--bhp->ref > 1 || (bhp->ref == 1 && !F_ISSET(bhp, BH_LOCKED))) {
                MUTEX_UNLOCK(dbenv, hp->mtx_hash);
                return (0);
        }

        /* Update priority values. */
        if (F_ISSET(bhp, BH_DISCARD) || mfp->priority == MPOOL_PRI_VERY_LOW)
                bhp->priority = 0;
        else {
                /*
                 * We don't lock the LRU counter or the stat.st_pages field, if
                 * we get garbage (which won't happen on a 32-bit machine), it
                 * only means a buffer has the wrong priority.
                 */
                bhp->priority = c_mp->lru_count;

                adjust = 0;
                if (mfp->priority != 0)
                        adjust =
                            (int)c_mp->stat.st_pages / mfp->priority;
                if (F_ISSET(bhp, BH_DIRTY))
                        adjust += c_mp->stat.st_pages / MPOOL_PRI_DIRTY;

                if (adjust > 0) {
                        if (UINT32_MAX - bhp->priority >= (u_int32_t)adjust)
                                bhp->priority += adjust;
                } else if (adjust < 0)
                        if (bhp->priority > (u_int32_t)-adjust)
                                bhp->priority += adjust;
        }

        /*
         * Buffers on hash buckets are sorted by priority -- move the buffer
         * to the correct position in the list.
         */
        if ((fbhp =
             SH_TAILQ_FIRST(&hp->hash_bucket, __bh)) ==
             SH_TAILQ_LAST(&hp->hash_bucket, hq, __bh))
                goto done;

        if (fbhp == bhp)
                fbhp = SH_TAILQ_NEXT(fbhp, hq, __bh);
        SH_TAILQ_REMOVE(&hp->hash_bucket, bhp, hq, __bh);

        for (prev = NULL; fbhp != NULL;
            prev = fbhp, fbhp = SH_TAILQ_NEXT(fbhp, hq, __bh))
                if (fbhp->priority > bhp->priority)
                        break;
        if (prev == NULL)
                SH_TAILQ_INSERT_HEAD(&hp->hash_bucket, bhp, hq, __bh);
        else
                SH_TAILQ_INSERT_AFTER(&hp->hash_bucket, prev, bhp, hq, __bh);

done:
        /* Reset the hash bucket's priority. */
        hp->hash_priority = SH_TAILQ_FIRSTP(&hp->hash_bucket, __bh)->priority;

#ifdef DIAGNOSTIC
        __memp_check_order(hp);
#endif

        /*
         * The sync code has a separate counter for buffers on which it waits.
         * It reads that value without holding a lock so we update it as the
         * last thing we do.  Once that value goes to 0, we won't see another
         * reference to that buffer being returned to the cache until the sync
         * code has finished, so we're safe as long as we don't let the value
         * go to 0 before we finish with the buffer.
         */
        if (F_ISSET(bhp, BH_LOCKED) && bhp->ref_sync != 0)
                --bhp->ref_sync;

        MUTEX_UNLOCK(dbenv, hp->mtx_hash);

        /*
         * On every buffer put we update the buffer generation number and check
         * for wraparound.
         */
        if (++c_mp->lru_count == UINT32_MAX)
                if ((t_ret =
                    __memp_reset_lru(dbenv, dbmp->reginfo)) != 0 && ret == 0)
                        ret = t_ret;

        return (ret);
}

/*
 * __memp_reset_lru --
 *      Reset the cache LRU counter.
 */
static int
__memp_reset_lru(dbenv, infop)
        DB_ENV *dbenv;
        REGINFO *infop;
{
        BH *bhp;
        DB_MPOOL_HASH *hp;
        MPOOL *c_mp;
        u_int32_t bucket;

        c_mp = infop->primary;

        /*
         * Update the counter so all future allocations will start at the
         * bottom.
         */
        c_mp->lru_count -= MPOOL_BASE_DECREMENT;

        /* Adjust the priority of every buffer in the system. */
        for (hp = R_ADDR(infop, c_mp->htab),
            bucket = 0; bucket < c_mp->htab_buckets; ++hp, ++bucket) {
                /*
                 * Skip empty buckets.
                 *
                 * We can check for empty buckets before locking as we
                 * only care if the pointer is zero or non-zero.
                 */
                if (SH_TAILQ_FIRST(&hp->hash_bucket, __bh) == NULL)
                        continue;

                MUTEX_LOCK(dbenv, hp->mtx_hash);
                for (bhp = SH_TAILQ_FIRST(&hp->hash_bucket, __bh);
                    bhp != NULL; bhp = SH_TAILQ_NEXT(bhp, hq, __bh))
                        if (bhp->priority != UINT32_MAX &&
                            bhp->priority > MPOOL_BASE_DECREMENT)
                                bhp->priority -= MPOOL_BASE_DECREMENT;
                MUTEX_UNLOCK(dbenv, hp->mtx_hash);
        }

        return (0);
}

---