dlmunlock.c

Go to the documentation of this file.

/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * dlmunlock.c
 *
 * underlying calls for unlocking locks
 *
 * Copyright (C) 2004 Oracle.  All rights reserved.
 *
 * 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 021110-1307, USA.
 *
 */


#include <linux/module.h>
#include <linux/fs.h>
#include <linux/types.h>
#include <linux/highmem.h>
#include <linux/init.h>
#include <linux/sysctl.h>
#include <linux/random.h>
#include <linux/blkdev.h>
#include <linux/socket.h>
#include <linux/inet.h>
#include <linux/spinlock.h>
#include <linux/delay.h>

#include "cluster/heartbeat.h"
#include "cluster/nodemanager.h"
#include "cluster/tcp.h"

#include "dlmapi.h"
#include "dlmcommon.h"

#define MLOG_MASK_PREFIX ML_DLM
#include "cluster/masklog.h"

#define DLM_UNLOCK_FREE_LOCK           0x00000001
#define DLM_UNLOCK_CALL_AST            0x00000002
#define DLM_UNLOCK_REMOVE_LOCK         0x00000004
#define DLM_UNLOCK_REGRANT_LOCK        0x00000008
#define DLM_UNLOCK_CLEAR_CONVERT_TYPE  0x00000010


static enum dlm_status dlm_get_cancel_actions(struct dlm_ctxt *dlm,
                          struct dlm_lock_resource *res,
                          struct dlm_lock *lock,
                          struct dlm_lockstatus *lksb,
                          int *actions);
static enum dlm_status dlm_get_unlock_actions(struct dlm_ctxt *dlm,
                          struct dlm_lock_resource *res,
                          struct dlm_lock *lock,
                          struct dlm_lockstatus *lksb,
                          int *actions);

static enum dlm_status dlm_send_remote_unlock_request(struct dlm_ctxt *dlm,
                         struct dlm_lock_resource *res,
                         struct dlm_lock *lock,
                         struct dlm_lockstatus *lksb,
                         int flags,
                         u8 owner);


/*
 * according to the spec:
 * http://opendlm.sourceforge.net/cvsmirror/opendlm/docs/dlmbook_final.pdf
 *
 *  flags & LKM_CANCEL != 0: must be converting or blocked
 *  flags & LKM_CANCEL == 0: must be granted
 *
 * So to unlock a converting lock, you must first cancel the
 * convert (passing LKM_CANCEL in flags), then call the unlock
 * again (with no LKM_CANCEL in flags).
 */


/*
 * locking:
 *   caller needs:  none
 *   taken:         res->spinlock and lock->spinlock taken and dropped
 *   held on exit:  none
 * returns: DLM_NORMAL, DLM_NOLOCKMGR, status from network
 * all callers should have taken an extra ref on lock coming in
 */
static enum dlm_status dlmunlock_common(struct dlm_ctxt *dlm,
                    struct dlm_lock_resource *res,
                    struct dlm_lock *lock,
                    struct dlm_lockstatus *lksb,
                    int flags, int *call_ast,
                    int master_node)
{
    enum dlm_status status;
    int actions = 0;
    int in_use;
        u8 owner;

    mlog(0, "master_node = %d, valblk = %d\n", master_node,
         flags & LKM_VALBLK);

    if (master_node)
        BUG_ON(res->owner != dlm->node_num);
    else
        BUG_ON(res->owner == dlm->node_num);

    spin_lock(&dlm->ast_lock);
    /* We want to be sure that we're not freeing a lock
     * that still has AST's pending... */
    in_use = !list_empty(&lock->ast_list);
    spin_unlock(&dlm->ast_lock);
    if (in_use && !(flags & LKM_CANCEL)) {
           mlog(ML_ERROR, "lockres %.*s: Someone is calling dlmunlock "
            "while waiting for an ast!", res->lockname.len,
            res->lockname.name);
        return DLM_BADPARAM;
    }

    spin_lock(&res->spinlock);
    if (res->state & DLM_LOCK_RES_IN_PROGRESS) {
        if (master_node && !(flags & LKM_CANCEL)) {
            mlog(ML_ERROR, "lockres in progress!\n");
            spin_unlock(&res->spinlock);
            return DLM_FORWARD;
        }
        /* ok for this to sleep if not in a network handler */
        __dlm_wait_on_lockres(res);
        res->state |= DLM_LOCK_RES_IN_PROGRESS;
    }
    spin_lock(&lock->spinlock);

    if (res->state & DLM_LOCK_RES_RECOVERING) {
        status = DLM_RECOVERING;
        goto leave;
    }

    if (res->state & DLM_LOCK_RES_MIGRATING) {
        status = DLM_MIGRATING;
        goto leave;
    }

    /* see above for what the spec says about
     * LKM_CANCEL and the lock queue state */
    if (flags & LKM_CANCEL)
        status = dlm_get_cancel_actions(dlm, res, lock, lksb, &actions);
    else
        status = dlm_get_unlock_actions(dlm, res, lock, lksb, &actions);

    if (status != DLM_NORMAL && (status != DLM_CANCELGRANT || !master_node))
        goto leave;

    /* By now this has been masked out of cancel requests. */
    if (flags & LKM_VALBLK) {
        /* make the final update to the lvb */
        if (master_node)
            memcpy(res->lvb, lksb->lvb, DLM_LVB_LEN);
        else
            flags |= LKM_PUT_LVB; /* let the send function
                           * handle it. */
    }

    if (!master_node) {
        owner = res->owner;
        /* drop locks and send message */
        if (flags & LKM_CANCEL)
            lock->cancel_pending = 1;
        else
            lock->unlock_pending = 1;
        spin_unlock(&lock->spinlock);
        spin_unlock(&res->spinlock);
        status = dlm_send_remote_unlock_request(dlm, res, lock, lksb,
                            flags, owner);
        spin_lock(&res->spinlock);
        spin_lock(&lock->spinlock);
        /* if the master told us the lock was already granted,
         * let the ast handle all of these actions */
        if (status == DLM_CANCELGRANT) {
            actions &= ~(DLM_UNLOCK_REMOVE_LOCK|
                     DLM_UNLOCK_REGRANT_LOCK|
                     DLM_UNLOCK_CLEAR_CONVERT_TYPE);
        } else if (status == DLM_RECOVERING ||
               status == DLM_MIGRATING ||
               status == DLM_FORWARD) {
            /* must clear the actions because this unlock
             * is about to be retried.  cannot free or do
             * any list manipulation. */
            mlog(0, "%s:%.*s: clearing actions, %s\n",
                 dlm->name, res->lockname.len,
                 res->lockname.name,
                 status==DLM_RECOVERING?"recovering":
                 (status==DLM_MIGRATING?"migrating":
                  "forward"));
            actions = 0;
        }
        if (flags & LKM_CANCEL)
            lock->cancel_pending = 0;
        else
            lock->unlock_pending = 0;

    }

    /* get an extra ref on lock.  if we are just switching
     * lists here, we dont want the lock to go away. */
    dlm_lock_get(lock);

    if (actions & DLM_UNLOCK_REMOVE_LOCK) {
        list_del_init(&lock->list);
        dlm_lock_put(lock);
    }
    if (actions & DLM_UNLOCK_REGRANT_LOCK) {
        dlm_lock_get(lock);
        list_add_tail(&lock->list, &res->granted);
    }
    if (actions & DLM_UNLOCK_CLEAR_CONVERT_TYPE) {
        mlog(0, "clearing convert_type at %smaster node\n",
             master_node ? "" : "non-");
        lock->ml.convert_type = LKM_IVMODE;
    }

    /* remove the extra ref on lock */
    dlm_lock_put(lock);

leave:
    res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
    if (!dlm_lock_on_list(&res->converting, lock))
        BUG_ON(lock->ml.convert_type != LKM_IVMODE);
    else
        BUG_ON(lock->ml.convert_type == LKM_IVMODE);
    spin_unlock(&lock->spinlock);
    spin_unlock(&res->spinlock);
    wake_up(&res->wq);

    /* let the caller's final dlm_lock_put handle the actual kfree */
    if (actions & DLM_UNLOCK_FREE_LOCK) {
        /* this should always be coupled with list removal */
        BUG_ON(!(actions & DLM_UNLOCK_REMOVE_LOCK));
        mlog(0, "lock %u:%llu should be gone now! refs=%d\n",
             dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
             dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
             atomic_read(&lock->lock_refs.refcount)-1);
        dlm_lock_put(lock);
    }
    if (actions & DLM_UNLOCK_CALL_AST)
        *call_ast = 1;

    /* if cancel or unlock succeeded, lvb work is done */
    if (status == DLM_NORMAL)
        lksb->flags &= ~(DLM_LKSB_PUT_LVB|DLM_LKSB_GET_LVB);

    return status;
}

void dlm_commit_pending_unlock(struct dlm_lock_resource *res,
                   struct dlm_lock *lock)
{
    /* leave DLM_LKSB_PUT_LVB on the lksb so any final
     * update of the lvb will be sent to the new master */
    list_del_init(&lock->list);
}

void dlm_commit_pending_cancel(struct dlm_lock_resource *res,
                   struct dlm_lock *lock)
{
    list_move_tail(&lock->list, &res->granted);
    lock->ml.convert_type = LKM_IVMODE;
}


static inline enum dlm_status dlmunlock_master(struct dlm_ctxt *dlm,
                      struct dlm_lock_resource *res,
                      struct dlm_lock *lock,
                      struct dlm_lockstatus *lksb,
                      int flags,
                      int *call_ast)
{
    return dlmunlock_common(dlm, res, lock, lksb, flags, call_ast, 1);
}

static inline enum dlm_status dlmunlock_remote(struct dlm_ctxt *dlm,
                      struct dlm_lock_resource *res,
                      struct dlm_lock *lock,
                      struct dlm_lockstatus *lksb,
                      int flags, int *call_ast)
{
    return dlmunlock_common(dlm, res, lock, lksb, flags, call_ast, 0);
}

/*
 * locking:
 *   caller needs:  none
 *   taken:         none
 *   held on exit:  none
 * returns: DLM_NORMAL, DLM_NOLOCKMGR, status from network
 */
static enum dlm_status dlm_send_remote_unlock_request(struct dlm_ctxt *dlm,
                         struct dlm_lock_resource *res,
                         struct dlm_lock *lock,
                         struct dlm_lockstatus *lksb,
                         int flags,
                         u8 owner)
{
    struct dlm_unlock_lock unlock;
    int tmpret;
    enum dlm_status ret;
    int status = 0;
    struct kvec vec[2];
    size_t veclen = 1;

    mlog(0, "%.*s\n", res->lockname.len, res->lockname.name);

    if (owner == dlm->node_num) {
        /* ended up trying to contact ourself.  this means
         * that the lockres had been remote but became local
         * via a migration.  just retry it, now as local */
        mlog(0, "%s:%.*s: this node became the master due to a "
             "migration, re-evaluate now\n", dlm->name,
             res->lockname.len, res->lockname.name);
        return DLM_FORWARD;
    }

    memset(&unlock, 0, sizeof(unlock));
    unlock.node_idx = dlm->node_num;
    unlock.flags = cpu_to_be32(flags);
    unlock.cookie = lock->ml.cookie;
    unlock.namelen = res->lockname.len;
    memcpy(unlock.name, res->lockname.name, unlock.namelen);

    vec[0].iov_len = sizeof(struct dlm_unlock_lock);
    vec[0].iov_base = &unlock;

    if (flags & LKM_PUT_LVB) {
        /* extra data to send if we are updating lvb */
        vec[1].iov_len = DLM_LVB_LEN;
        vec[1].iov_base = lock->lksb->lvb;
        veclen++;
    }

    tmpret = o2net_send_message_vec(DLM_UNLOCK_LOCK_MSG, dlm->key,
                    vec, veclen, owner, &status);
    if (tmpret >= 0) {
        // successfully sent and received
        if (status == DLM_FORWARD)
            mlog(0, "master was in-progress.  retry\n");
        ret = status;
    } else {
        mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
             "node %u\n", tmpret, DLM_UNLOCK_LOCK_MSG, dlm->key, owner);
        if (dlm_is_host_down(tmpret)) {
            /* NOTE: this seems strange, but it is what we want.
             * when the master goes down during a cancel or
             * unlock, the recovery code completes the operation
             * as if the master had not died, then passes the
             * updated state to the recovery master.  this thread
             * just needs to finish out the operation and call
             * the unlockast. */
            ret = DLM_NORMAL;
        } else {
            /* something bad.  this will BUG in ocfs2 */
            ret = dlm_err_to_dlm_status(tmpret);
        }
    }

    return ret;
}

/*
 * locking:
 *   caller needs:  none
 *   taken:         takes and drops res->spinlock
 *   held on exit:  none
 * returns: DLM_NORMAL, DLM_BADARGS, DLM_IVLOCKID,
 *          return value from dlmunlock_master
 */
int dlm_unlock_lock_handler(struct o2net_msg *msg, u32 len, void *data,
                void **ret_data)
{
    struct dlm_ctxt *dlm = data;
    struct dlm_unlock_lock *unlock = (struct dlm_unlock_lock *)msg->buf;
    struct dlm_lock_resource *res = NULL;
    struct list_head *iter;
    struct dlm_lock *lock = NULL;
    enum dlm_status status = DLM_NORMAL;
    int found = 0, i;
    struct dlm_lockstatus *lksb = NULL;
    int ignore;
    u32 flags;
    struct list_head *queue;

    flags = be32_to_cpu(unlock->flags);

    if (flags & LKM_GET_LVB) {
        mlog(ML_ERROR, "bad args!  GET_LVB specified on unlock!\n");
        return DLM_BADARGS;
    }

    if ((flags & (LKM_PUT_LVB|LKM_CANCEL)) == (LKM_PUT_LVB|LKM_CANCEL)) {
        mlog(ML_ERROR, "bad args!  cannot modify lvb on a CANCEL "
             "request!\n");
        return DLM_BADARGS;
    }

    if (unlock->namelen > DLM_LOCKID_NAME_MAX) {
        mlog(ML_ERROR, "Invalid name length in unlock handler!\n");
        return DLM_IVBUFLEN;
    }

    if (!dlm_grab(dlm))
        return DLM_REJECTED;

    mlog_bug_on_msg(!dlm_domain_fully_joined(dlm),
            "Domain %s not fully joined!\n", dlm->name);

    mlog(0, "lvb: %s\n", flags & LKM_PUT_LVB ? "put lvb" : "none");

    res = dlm_lookup_lockres(dlm, unlock->name, unlock->namelen);
    if (!res) {
        /* We assume here that a no lock resource simply means
         * it was migrated away and destroyed before the other
         * node could detect it. */
        mlog(0, "returning DLM_FORWARD -- res no longer exists\n");
        status = DLM_FORWARD;
        goto not_found;
    }

    queue=&res->granted;
    found = 0;
    spin_lock(&res->spinlock);
    if (res->state & DLM_LOCK_RES_RECOVERING) {
        spin_unlock(&res->spinlock);
        mlog(0, "returning DLM_RECOVERING\n");
        status = DLM_RECOVERING;
        goto leave;
    }

    if (res->state & DLM_LOCK_RES_MIGRATING) {
        spin_unlock(&res->spinlock);
        mlog(0, "returning DLM_MIGRATING\n");
        status = DLM_MIGRATING;
        goto leave;
    }

    if (res->owner != dlm->node_num) {
        spin_unlock(&res->spinlock);
        mlog(0, "returning DLM_FORWARD -- not master\n");
        status = DLM_FORWARD;
        goto leave;
    }

    for (i=0; i<3; i++) {
        list_for_each(iter, queue) {
            lock = list_entry(iter, struct dlm_lock, list);
            if (lock->ml.cookie == unlock->cookie &&
                    lock->ml.node == unlock->node_idx) {
                dlm_lock_get(lock);
                found = 1;
                break;
            }
        }
        if (found)
            break;
        /* scan granted -> converting -> blocked queues */
        queue++;
    }
    spin_unlock(&res->spinlock);
    if (!found) {
        status = DLM_IVLOCKID;
        goto not_found;
    }

    /* lock was found on queue */
    lksb = lock->lksb;
    if (flags & (LKM_VALBLK|LKM_PUT_LVB) &&
        lock->ml.type != LKM_EXMODE)
        flags &= ~(LKM_VALBLK|LKM_PUT_LVB);

    /* unlockast only called on originating node */
    if (flags & LKM_PUT_LVB) {
        lksb->flags |= DLM_LKSB_PUT_LVB;
        memcpy(&lksb->lvb[0], &unlock->lvb[0], DLM_LVB_LEN);
    }

    /* if this is in-progress, propagate the DLM_FORWARD
     * all the way back out */
    status = dlmunlock_master(dlm, res, lock, lksb, flags, &ignore);
    if (status == DLM_FORWARD)
        mlog(0, "lockres is in progress\n");

    if (flags & LKM_PUT_LVB)
        lksb->flags &= ~DLM_LKSB_PUT_LVB;

    dlm_lockres_calc_usage(dlm, res);
    dlm_kick_thread(dlm, res);

not_found:
    if (!found)
        mlog(ML_ERROR, "failed to find lock to unlock! "
                   "cookie=%u:%llu\n",
             dlm_get_lock_cookie_node(be64_to_cpu(unlock->cookie)),
             dlm_get_lock_cookie_seq(be64_to_cpu(unlock->cookie)));
    else
        dlm_lock_put(lock);

leave:
    if (res)
        dlm_lockres_put(res);

    dlm_put(dlm);

    return status;
}


static enum dlm_status dlm_get_cancel_actions(struct dlm_ctxt *dlm,
                          struct dlm_lock_resource *res,
                          struct dlm_lock *lock,
                          struct dlm_lockstatus *lksb,
                          int *actions)
{
    enum dlm_status status;

    if (dlm_lock_on_list(&res->blocked, lock)) {
        /* cancel this outright */
        status = DLM_NORMAL;
        *actions = (DLM_UNLOCK_CALL_AST |
                DLM_UNLOCK_REMOVE_LOCK);
    } else if (dlm_lock_on_list(&res->converting, lock)) {
        /* cancel the request, put back on granted */
        status = DLM_NORMAL;
        *actions = (DLM_UNLOCK_CALL_AST |
                DLM_UNLOCK_REMOVE_LOCK |
                DLM_UNLOCK_REGRANT_LOCK |
                DLM_UNLOCK_CLEAR_CONVERT_TYPE);
    } else if (dlm_lock_on_list(&res->granted, lock)) {
        /* too late, already granted. */
        status = DLM_CANCELGRANT;
        *actions = DLM_UNLOCK_CALL_AST;
    } else {
        mlog(ML_ERROR, "lock to cancel is not on any list!\n");
        status = DLM_IVLOCKID;
        *actions = 0;
    }
    return status;
}

static enum dlm_status dlm_get_unlock_actions(struct dlm_ctxt *dlm,
                          struct dlm_lock_resource *res,
                          struct dlm_lock *lock,
                          struct dlm_lockstatus *lksb,
                          int *actions)
{
    enum dlm_status status;

    /* unlock request */
    if (!dlm_lock_on_list(&res->granted, lock)) {
        status = DLM_DENIED;
        dlm_error(status);
        *actions = 0;
    } else {
        /* unlock granted lock */
        status = DLM_NORMAL;
        *actions = (DLM_UNLOCK_FREE_LOCK |
                DLM_UNLOCK_CALL_AST |
                DLM_UNLOCK_REMOVE_LOCK);
    }
    return status;
}

/* there seems to be no point in doing this async
 * since (even for the remote case) there is really
 * no work to queue up... so just do it and fire the
 * unlockast by hand when done... */
enum dlm_status dlmunlock(struct dlm_ctxt *dlm, struct dlm_lockstatus *lksb,
              int flags, dlm_astunlockfunc_t *unlockast, void *data)
{
    enum dlm_status status;
    struct dlm_lock_resource *res;
    struct dlm_lock *lock = NULL;
    int call_ast, is_master;

    if (!lksb) {
        dlm_error(DLM_BADARGS);
        return DLM_BADARGS;
    }

    if (flags & ~(LKM_CANCEL | LKM_VALBLK | LKM_INVVALBLK)) {
        dlm_error(DLM_BADPARAM);
        return DLM_BADPARAM;
    }

    if ((flags & (LKM_VALBLK | LKM_CANCEL)) == (LKM_VALBLK | LKM_CANCEL)) {
        mlog(0, "VALBLK given with CANCEL: ignoring VALBLK\n");
        flags &= ~LKM_VALBLK;
    }

    if (!lksb->lockid || !lksb->lockid->lockres) {
        dlm_error(DLM_BADPARAM);
        return DLM_BADPARAM;
    }

    lock = lksb->lockid;
    BUG_ON(!lock);
    dlm_lock_get(lock);

    res = lock->lockres;
    BUG_ON(!res);
    dlm_lockres_get(res);
retry:
    call_ast = 0;
    /* need to retry up here because owner may have changed */
    mlog(0, "lock=%p res=%p\n", lock, res);

    spin_lock(&res->spinlock);
    is_master = (res->owner == dlm->node_num);
    if (flags & LKM_VALBLK && lock->ml.type != LKM_EXMODE)
        flags &= ~LKM_VALBLK;
    spin_unlock(&res->spinlock);

    if (is_master) {
        status = dlmunlock_master(dlm, res, lock, lksb, flags,
                      &call_ast);
        mlog(0, "done calling dlmunlock_master: returned %d, "
             "call_ast is %d\n", status, call_ast);
    } else {
        status = dlmunlock_remote(dlm, res, lock, lksb, flags,
                      &call_ast);
        mlog(0, "done calling dlmunlock_remote: returned %d, "
             "call_ast is %d\n", status, call_ast);
    }

    if (status == DLM_RECOVERING ||
        status == DLM_MIGRATING ||
        status == DLM_FORWARD) {
        /* We want to go away for a tiny bit to allow recovery
         * / migration to complete on this resource. I don't
         * know of any wait queue we could sleep on as this
         * may be happening on another node. Perhaps the
         * proper solution is to queue up requests on the
         * other end? */

        /* do we want to yield(); ?? */
        msleep(50);

        mlog(0, "retrying unlock due to pending recovery/"
             "migration/in-progress\n");
        goto retry;
    }

    if (call_ast) {
        mlog(0, "calling unlockast(%p, %d)\n", data, status);
        if (is_master) {
            /* it is possible that there is one last bast
             * pending.  make sure it is flushed, then
             * call the unlockast.
             * not an issue if this is a mastered remotely,
             * since this lock has been removed from the
             * lockres queues and cannot be found. */
            dlm_kick_thread(dlm, NULL);
            wait_event(dlm->ast_wq,
                   dlm_lock_basts_flushed(dlm, lock));
        }
        (*unlockast)(data, status);
    }

    if (status == DLM_CANCELGRANT)
        status = DLM_NORMAL;

    if (status == DLM_NORMAL) {
        mlog(0, "kicking the thread\n");
        dlm_kick_thread(dlm, res);
    } else
        dlm_error(status);

    dlm_lockres_calc_usage(dlm, res);
    dlm_lockres_put(res);
    dlm_lock_put(lock);

    mlog(0, "returning status=%d!\n", status);
    return status;
}
EXPORT_SYMBOL_GPL(dlmunlock);