[linux-block.git] / fs / dlm / rcom.c

// SPDX-License-Identifier: GPL-2.0-only
/******************************************************************************
*******************************************************************************
**
**  Copyright (C) Sistina Software, Inc.  1997-2003  All rights reserved.
**  Copyright (C) 2005-2008 Red Hat, Inc.  All rights reserved.
**
**
*******************************************************************************
******************************************************************************/

#include "dlm_internal.h"
#include "lockspace.h"
#include "member.h"
#include "lowcomms.h"
#include "midcomms.h"
#include "rcom.h"
#include "recover.h"
#include "dir.h"
#include "config.h"
#include "memory.h"
#include "lock.h"
#include "util.h"

static int rcom_response(struct dlm_ls *ls)
{
	return test_bit(LSFL_RCOM_READY, &ls->ls_flags);
}

static void _create_rcom(struct dlm_ls *ls, int to_nodeid, int type, int len,
			 struct dlm_rcom **rc_ret, char *mb, int mb_len)
{
	struct dlm_rcom *rc;

	rc = (struct dlm_rcom *) mb;

	rc->rc_header.h_version = (DLM_HEADER_MAJOR | DLM_HEADER_MINOR);
	rc->rc_header.u.h_lockspace = ls->ls_global_id;
	rc->rc_header.h_nodeid = dlm_our_nodeid();
	rc->rc_header.h_length = mb_len;
	rc->rc_header.h_cmd = DLM_RCOM;

	rc->rc_type = type;

	spin_lock(&ls->ls_recover_lock);
	rc->rc_seq = ls->ls_recover_seq;
	spin_unlock(&ls->ls_recover_lock);

	*rc_ret = rc;
}

static int create_rcom(struct dlm_ls *ls, int to_nodeid, int type, int len,
		       struct dlm_rcom **rc_ret, struct dlm_mhandle **mh_ret)
{
	int mb_len = sizeof(struct dlm_rcom) + len;
	struct dlm_mhandle *mh;
	char *mb;

	mh = dlm_midcomms_get_mhandle(to_nodeid, mb_len, GFP_NOFS, &mb);
	if (!mh) {
		log_print("%s to %d type %d len %d ENOBUFS",
			  __func__, to_nodeid, type, len);
		return -ENOBUFS;
	}

	_create_rcom(ls, to_nodeid, type, len, rc_ret, mb, mb_len);
	*mh_ret = mh;
	return 0;
}

static int create_rcom_stateless(struct dlm_ls *ls, int to_nodeid, int type,
				 int len, struct dlm_rcom **rc_ret,
				 struct dlm_msg **msg_ret)
{
	int mb_len = sizeof(struct dlm_rcom) + len;
	struct dlm_msg *msg;
	char *mb;

	msg = dlm_lowcomms_new_msg(to_nodeid, mb_len, GFP_NOFS, &mb,
				   NULL, NULL);
	if (!msg) {
		log_print("create_rcom to %d type %d len %d ENOBUFS",
			  to_nodeid, type, len);
		return -ENOBUFS;
	}

	_create_rcom(ls, to_nodeid, type, len, rc_ret, mb, mb_len);
	*msg_ret = msg;
	return 0;
}

static void _send_rcom(struct dlm_ls *ls, struct dlm_rcom *rc)
{
	dlm_rcom_out(rc);
}

static void send_rcom(struct dlm_ls *ls, struct dlm_mhandle *mh,
		      struct dlm_rcom *rc)
{
	_send_rcom(ls, rc);
	dlm_midcomms_commit_mhandle(mh);
}

static void send_rcom_stateless(struct dlm_ls *ls, struct dlm_msg *msg,
				struct dlm_rcom *rc)
{
	_send_rcom(ls, rc);
	dlm_lowcomms_commit_msg(msg);
	dlm_lowcomms_put_msg(msg);
}

static void set_rcom_status(struct dlm_ls *ls, struct rcom_status *rs,
			    uint32_t flags)
{
	rs->rs_flags = cpu_to_le32(flags);
}

/* When replying to a status request, a node also sends back its
   configuration values.  The requesting node then checks that the remote
   node is configured the same way as itself. */

static void set_rcom_config(struct dlm_ls *ls, struct rcom_config *rf,
			    uint32_t num_slots)
{
	rf->rf_lvblen = cpu_to_le32(ls->ls_lvblen);
	rf->rf_lsflags = cpu_to_le32(ls->ls_exflags);

	rf->rf_our_slot = cpu_to_le16(ls->ls_slot);
	rf->rf_num_slots = cpu_to_le16(num_slots);
	rf->rf_generation =  cpu_to_le32(ls->ls_generation);
}

static int check_rcom_config(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid)
{
	struct rcom_config *rf = (struct rcom_config *) rc->rc_buf;

	if ((rc->rc_header.h_version & 0xFFFF0000) != DLM_HEADER_MAJOR) {
		log_error(ls, "version mismatch: %x nodeid %d: %x",
			  DLM_HEADER_MAJOR | DLM_HEADER_MINOR, nodeid,
			  rc->rc_header.h_version);
		return -EPROTO;
	}

	if (le32_to_cpu(rf->rf_lvblen) != ls->ls_lvblen ||
	    le32_to_cpu(rf->rf_lsflags) != ls->ls_exflags) {
		log_error(ls, "config mismatch: %d,%x nodeid %d: %d,%x",
			  ls->ls_lvblen, ls->ls_exflags, nodeid,
			  le32_to_cpu(rf->rf_lvblen),
			  le32_to_cpu(rf->rf_lsflags));
		return -EPROTO;
	}
	return 0;
}

static void allow_sync_reply(struct dlm_ls *ls, uint64_t *new_seq)
{
	spin_lock(&ls->ls_rcom_spin);
	*new_seq = ++ls->ls_rcom_seq;
	set_bit(LSFL_RCOM_WAIT, &ls->ls_flags);
	spin_unlock(&ls->ls_rcom_spin);
}

static void disallow_sync_reply(struct dlm_ls *ls)
{
	spin_lock(&ls->ls_rcom_spin);
	clear_bit(LSFL_RCOM_WAIT, &ls->ls_flags);
	clear_bit(LSFL_RCOM_READY, &ls->ls_flags);
	spin_unlock(&ls->ls_rcom_spin);
}

/*
 * low nodeid gathers one slot value at a time from each node.
 * it sets need_slots=0, and saves rf_our_slot returned from each
 * rcom_config.
 *
 * other nodes gather all slot values at once from the low nodeid.
 * they set need_slots=1, and ignore the rf_our_slot returned from each
 * rcom_config.  they use the rf_num_slots returned from the low
 * node's rcom_config.
 */

int dlm_rcom_status(struct dlm_ls *ls, int nodeid, uint32_t status_flags)
{
	struct dlm_rcom *rc;
	struct dlm_msg *msg;
	int error = 0;

	ls->ls_recover_nodeid = nodeid;

	if (nodeid == dlm_our_nodeid()) {
		rc = ls->ls_recover_buf;
		rc->rc_result = dlm_recover_status(ls);
		goto out;
	}

retry:
	error = create_rcom_stateless(ls, nodeid, DLM_RCOM_STATUS,
				      sizeof(struct rcom_status), &rc, &msg);
	if (error)
		goto out;

	set_rcom_status(ls, (struct rcom_status *)rc->rc_buf, status_flags);

	allow_sync_reply(ls, &rc->rc_id);
	memset(ls->ls_recover_buf, 0, DEFAULT_BUFFER_SIZE);

	send_rcom_stateless(ls, msg, rc);

	error = dlm_wait_function(ls, &rcom_response);
	disallow_sync_reply(ls);
	if (error == -ETIMEDOUT)
		goto retry;
	if (error)
		goto out;

	rc = ls->ls_recover_buf;

	if (rc->rc_result == -ESRCH) {
		/* we pretend the remote lockspace exists with 0 status */
		log_debug(ls, "remote node %d not ready", nodeid);
		rc->rc_result = 0;
		error = 0;
	} else {
		error = check_rcom_config(ls, rc, nodeid);
	}

	/* the caller looks at rc_result for the remote recovery status */
 out:
	return error;
}

static void receive_rcom_status(struct dlm_ls *ls, struct dlm_rcom *rc_in)
{
	struct dlm_rcom *rc;
	struct rcom_status *rs;
	uint32_t status;
	int nodeid = rc_in->rc_header.h_nodeid;
	int len = sizeof(struct rcom_config);
	struct dlm_msg *msg;
	int num_slots = 0;
	int error;

	if (!dlm_slots_version(&rc_in->rc_header)) {
		status = dlm_recover_status(ls);
		goto do_create;
	}

	rs = (struct rcom_status *)rc_in->rc_buf;

	if (!(le32_to_cpu(rs->rs_flags) & DLM_RSF_NEED_SLOTS)) {
		status = dlm_recover_status(ls);
		goto do_create;
	}

	spin_lock(&ls->ls_recover_lock);
	status = ls->ls_recover_status;
	num_slots = ls->ls_num_slots;
	spin_unlock(&ls->ls_recover_lock);
	len += num_slots * sizeof(struct rcom_slot);

 do_create:
	error = create_rcom_stateless(ls, nodeid, DLM_RCOM_STATUS_REPLY,
				      len, &rc, &msg);
	if (error)
		return;

	rc->rc_id = rc_in->rc_id;
	rc->rc_seq_reply = rc_in->rc_seq;
	rc->rc_result = status;

	set_rcom_config(ls, (struct rcom_config *)rc->rc_buf, num_slots);

	if (!num_slots)
		goto do_send;

	spin_lock(&ls->ls_recover_lock);
	if (ls->ls_num_slots != num_slots) {
		spin_unlock(&ls->ls_recover_lock);
		log_debug(ls, "receive_rcom_status num_slots %d to %d",
			  num_slots, ls->ls_num_slots);
		rc->rc_result = 0;
		set_rcom_config(ls, (struct rcom_config *)rc->rc_buf, 0);
		goto do_send;
	}

	dlm_slots_copy_out(ls, rc);
	spin_unlock(&ls->ls_recover_lock);

 do_send:
	send_rcom_stateless(ls, msg, rc);
}

static void receive_sync_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
{
	spin_lock(&ls->ls_rcom_spin);
	if (!test_bit(LSFL_RCOM_WAIT, &ls->ls_flags) ||
	    rc_in->rc_id != ls->ls_rcom_seq) {
		log_debug(ls, "reject reply %d from %d seq %llx expect %llx",
			  rc_in->rc_type, rc_in->rc_header.h_nodeid,
			  (unsigned long long)rc_in->rc_id,
			  (unsigned long long)ls->ls_rcom_seq);
		goto out;
	}
	memcpy(ls->ls_recover_buf, rc_in, rc_in->rc_header.h_length);
	set_bit(LSFL_RCOM_READY, &ls->ls_flags);
	clear_bit(LSFL_RCOM_WAIT, &ls->ls_flags);
	wake_up(&ls->ls_wait_general);
 out:
	spin_unlock(&ls->ls_rcom_spin);
}

int dlm_rcom_names(struct dlm_ls *ls, int nodeid, char *last_name, int last_len)
{
	struct dlm_rcom *rc;
	struct dlm_msg *msg;
	int error = 0;

	ls->ls_recover_nodeid = nodeid;

retry:
	error = create_rcom_stateless(ls, nodeid, DLM_RCOM_NAMES, last_len,
				      &rc, &msg);
	if (error)
		goto out;
	memcpy(rc->rc_buf, last_name, last_len);

	allow_sync_reply(ls, &rc->rc_id);
	memset(ls->ls_recover_buf, 0, DEFAULT_BUFFER_SIZE);

	send_rcom_stateless(ls, msg, rc);

	error = dlm_wait_function(ls, &rcom_response);
	disallow_sync_reply(ls);
	if (error == -ETIMEDOUT)
		goto retry;
 out:
	return error;
}

static void receive_rcom_names(struct dlm_ls *ls, struct dlm_rcom *rc_in)
{
	struct dlm_rcom *rc;
	int error, inlen, outlen, nodeid;
	struct dlm_msg *msg;

	nodeid = rc_in->rc_header.h_nodeid;
	inlen = rc_in->rc_header.h_length - sizeof(struct dlm_rcom);
	outlen = LOWCOMMS_MAX_TX_BUFFER_LEN - sizeof(struct dlm_rcom);

	error = create_rcom_stateless(ls, nodeid, DLM_RCOM_NAMES_REPLY, outlen,
				      &rc, &msg);
	if (error)
		return;
	rc->rc_id = rc_in->rc_id;
	rc->rc_seq_reply = rc_in->rc_seq;

	dlm_copy_master_names(ls, rc_in->rc_buf, inlen, rc->rc_buf, outlen,
			      nodeid);
	send_rcom_stateless(ls, msg, rc);
}

int dlm_send_rcom_lookup(struct dlm_rsb *r, int dir_nodeid)
{
	struct dlm_rcom *rc;
	struct dlm_mhandle *mh;
	struct dlm_ls *ls = r->res_ls;
	int error;

	error = create_rcom(ls, dir_nodeid, DLM_RCOM_LOOKUP, r->res_length,
			    &rc, &mh);
	if (error)
		goto out;
	memcpy(rc->rc_buf, r->res_name, r->res_length);
	rc->rc_id = (unsigned long) r->res_id;

	send_rcom(ls, mh, rc);
 out:
	return error;
}

static void receive_rcom_lookup(struct dlm_ls *ls, struct dlm_rcom *rc_in)
{
	struct dlm_rcom *rc;
	struct dlm_mhandle *mh;
	int error, ret_nodeid, nodeid = rc_in->rc_header.h_nodeid;
	int len = rc_in->rc_header.h_length - sizeof(struct dlm_rcom);

	/* Old code would send this special id to trigger a debug dump. */
	if (rc_in->rc_id == 0xFFFFFFFF) {
		log_error(ls, "receive_rcom_lookup dump from %d", nodeid);
		dlm_dump_rsb_name(ls, rc_in->rc_buf, len);
		return;
	}

	error = create_rcom(ls, nodeid, DLM_RCOM_LOOKUP_REPLY, 0, &rc, &mh);
	if (error)
		return;

	error = dlm_master_lookup(ls, nodeid, rc_in->rc_buf, len,
				  DLM_LU_RECOVER_MASTER, &ret_nodeid, NULL);
	if (error)
		ret_nodeid = error;
	rc->rc_result = ret_nodeid;
	rc->rc_id = rc_in->rc_id;
	rc->rc_seq_reply = rc_in->rc_seq;

	send_rcom(ls, mh, rc);
}

static void receive_rcom_lookup_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
{
	dlm_recover_master_reply(ls, rc_in);
}

static void pack_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb,
			   struct rcom_lock *rl)
{
	memset(rl, 0, sizeof(*rl));

	rl->rl_ownpid = cpu_to_le32(lkb->lkb_ownpid);
	rl->rl_lkid = cpu_to_le32(lkb->lkb_id);
	rl->rl_exflags = cpu_to_le32(lkb->lkb_exflags);
	rl->rl_flags = cpu_to_le32(lkb->lkb_flags);
	rl->rl_lvbseq = cpu_to_le32(lkb->lkb_lvbseq);
	rl->rl_rqmode = lkb->lkb_rqmode;
	rl->rl_grmode = lkb->lkb_grmode;
	rl->rl_status = lkb->lkb_status;
	rl->rl_wait_type = cpu_to_le16(lkb->lkb_wait_type);

	if (lkb->lkb_bastfn)
		rl->rl_asts |= DLM_CB_BAST;
	if (lkb->lkb_astfn)
		rl->rl_asts |= DLM_CB_CAST;

	rl->rl_namelen = cpu_to_le16(r->res_length);
	memcpy(rl->rl_name, r->res_name, r->res_length);

	/* FIXME: might we have an lvb without DLM_LKF_VALBLK set ?
	   If so, receive_rcom_lock_args() won't take this copy. */

	if (lkb->lkb_lvbptr)
		memcpy(rl->rl_lvb, lkb->lkb_lvbptr, r->res_ls->ls_lvblen);
}

int dlm_send_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb)
{
	struct dlm_ls *ls = r->res_ls;
	struct dlm_rcom *rc;
	struct dlm_mhandle *mh;
	struct rcom_lock *rl;
	int error, len = sizeof(struct rcom_lock);

	if (lkb->lkb_lvbptr)
		len += ls->ls_lvblen;

	error = create_rcom(ls, r->res_nodeid, DLM_RCOM_LOCK, len, &rc, &mh);
	if (error)
		goto out;

	rl = (struct rcom_lock *) rc->rc_buf;
	pack_rcom_lock(r, lkb, rl);
	rc->rc_id = (unsigned long) r;

	send_rcom(ls, mh, rc);
 out:
	return error;
}

/* needs at least dlm_rcom + rcom_lock */
static void receive_rcom_lock(struct dlm_ls *ls, struct dlm_rcom *rc_in)
{
	struct dlm_rcom *rc;
	struct dlm_mhandle *mh;
	int error, nodeid = rc_in->rc_header.h_nodeid;

	dlm_recover_master_copy(ls, rc_in);

	error = create_rcom(ls, nodeid, DLM_RCOM_LOCK_REPLY,
			    sizeof(struct rcom_lock), &rc, &mh);
	if (error)
		return;

	/* We send back the same rcom_lock struct we received, but
	   dlm_recover_master_copy() has filled in rl_remid and rl_result */

	memcpy(rc->rc_buf, rc_in->rc_buf, sizeof(struct rcom_lock));
	rc->rc_id = rc_in->rc_id;
	rc->rc_seq_reply = rc_in->rc_seq;

	send_rcom(ls, mh, rc);
}

/* If the lockspace doesn't exist then still send a status message
   back; it's possible that it just doesn't have its global_id yet. */

int dlm_send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in)
{
	struct dlm_rcom *rc;
	struct rcom_config *rf;
	struct dlm_mhandle *mh;
	char *mb;
	int mb_len = sizeof(struct dlm_rcom) + sizeof(struct rcom_config);

	mh = dlm_midcomms_get_mhandle(nodeid, mb_len, GFP_NOFS, &mb);
	if (!mh)
		return -ENOBUFS;

	rc = (struct dlm_rcom *) mb;

	rc->rc_header.h_version = (DLM_HEADER_MAJOR | DLM_HEADER_MINOR);
	rc->rc_header.u.h_lockspace = rc_in->rc_header.u.h_lockspace;
	rc->rc_header.h_nodeid = dlm_our_nodeid();
	rc->rc_header.h_length = mb_len;
	rc->rc_header.h_cmd = DLM_RCOM;

	rc->rc_type = DLM_RCOM_STATUS_REPLY;
	rc->rc_id = rc_in->rc_id;
	rc->rc_seq_reply = rc_in->rc_seq;
	rc->rc_result = -ESRCH;

	rf = (struct rcom_config *) rc->rc_buf;
	rf->rf_lvblen = cpu_to_le32(~0U);

	dlm_rcom_out(rc);
	dlm_midcomms_commit_mhandle(mh);

	return 0;
}

/*
 * Ignore messages for stage Y before we set
 * recover_status bit for stage X:
 *
 * recover_status = 0
 *
 * dlm_recover_members()
 * - send nothing
 * - recv nothing
 * - ignore NAMES, NAMES_REPLY
 * - ignore LOOKUP, LOOKUP_REPLY
 * - ignore LOCK, LOCK_REPLY
 *
 * recover_status |= NODES
 *
 * dlm_recover_members_wait()
 *
 * dlm_recover_directory()
 * - send NAMES
 * - recv NAMES_REPLY
 * - ignore LOOKUP, LOOKUP_REPLY
 * - ignore LOCK, LOCK_REPLY
 *
 * recover_status |= DIR
 *
 * dlm_recover_directory_wait()
 *
 * dlm_recover_masters()
 * - send LOOKUP
 * - recv LOOKUP_REPLY
 *
 * dlm_recover_locks()
 * - send LOCKS
 * - recv LOCKS_REPLY
 *
 * recover_status |= LOCKS
 *
 * dlm_recover_locks_wait()
 *
 * recover_status |= DONE
 */

/* Called by dlm_recv; corresponds to dlm_receive_message() but special
   recovery-only comms are sent through here. */

void dlm_receive_rcom(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid)
{
	int lock_size = sizeof(struct dlm_rcom) + sizeof(struct rcom_lock);
	int stop, reply = 0, names = 0, lookup = 0, lock = 0;
	uint32_t status;
	uint64_t seq;

	switch (rc->rc_type) {
	case DLM_RCOM_STATUS_REPLY:
		reply = 1;
		break;
	case DLM_RCOM_NAMES:
		names = 1;
		break;
	case DLM_RCOM_NAMES_REPLY:
		names = 1;
		reply = 1;
		break;
	case DLM_RCOM_LOOKUP:
		lookup = 1;
		break;
	case DLM_RCOM_LOOKUP_REPLY:
		lookup = 1;
		reply = 1;
		break;
	case DLM_RCOM_LOCK:
		lock = 1;
		break;
	case DLM_RCOM_LOCK_REPLY:
		lock = 1;
		reply = 1;
		break;
	}

	spin_lock(&ls->ls_recover_lock);
	status = ls->ls_recover_status;
	stop = test_bit(LSFL_RECOVER_STOP, &ls->ls_flags);
	seq = ls->ls_recover_seq;
	spin_unlock(&ls->ls_recover_lock);

	if (stop && (rc->rc_type != DLM_RCOM_STATUS))
		goto ignore;

	if (reply && (rc->rc_seq_reply != seq))
		goto ignore;

	if (!(status & DLM_RS_NODES) && (names || lookup || lock))
		goto ignore;

	if (!(status & DLM_RS_DIR) && (lookup || lock))
		goto ignore;

	switch (rc->rc_type) {
	case DLM_RCOM_STATUS:
		receive_rcom_status(ls, rc);
		break;

	case DLM_RCOM_NAMES:
		receive_rcom_names(ls, rc);
		break;

	case DLM_RCOM_LOOKUP:
		receive_rcom_lookup(ls, rc);
		break;

	case DLM_RCOM_LOCK:
		if (rc->rc_header.h_length < lock_size)
			goto Eshort;
		receive_rcom_lock(ls, rc);
		break;

	case DLM_RCOM_STATUS_REPLY:
		receive_sync_reply(ls, rc);
		break;

	case DLM_RCOM_NAMES_REPLY:
		receive_sync_reply(ls, rc);
		break;

	case DLM_RCOM_LOOKUP_REPLY:
		receive_rcom_lookup_reply(ls, rc);
		break;

	case DLM_RCOM_LOCK_REPLY:
		if (rc->rc_header.h_length < lock_size)
			goto Eshort;
		dlm_recover_process_copy(ls, rc);
		break;

	default:
		log_error(ls, "receive_rcom bad type %d", rc->rc_type);
	}
	return;

ignore:
	log_limit(ls, "dlm_receive_rcom ignore msg %d "
		  "from %d %llu %llu recover seq %llu sts %x gen %u",
		   rc->rc_type,
		   nodeid,
		   (unsigned long long)rc->rc_seq,
		   (unsigned long long)rc->rc_seq_reply,
		   (unsigned long long)seq,
		   status, ls->ls_generation);
	return;
Eshort:
	log_error(ls, "recovery message %d from %d is too short",
		  rc->rc_type, nodeid);
}
Commit	Line	Data
2522fe45	1	// SPDX-License-Identifier: GPL-2.0-only
e7fd4179 DT	2	/******************************************************************************
	3	*******************************************************************************
	4	**
	5	** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
dbcfc347	6	** Copyright (C) 2005-2008 Red Hat, Inc. All rights reserved.
e7fd4179	7	**
e7fd4179 DT	8	**
	9	*******************************************************************************
	10	******************************************************************************/
	11
	12	#include "dlm_internal.h"
	13	#include "lockspace.h"
	14	#include "member.h"
	15	#include "lowcomms.h"
	16	#include "midcomms.h"
	17	#include "rcom.h"
	18	#include "recover.h"
	19	#include "dir.h"
	20	#include "config.h"
	21	#include "memory.h"
	22	#include "lock.h"
	23	#include "util.h"
e7fd4179 DT	24
	25	static int rcom_response(struct dlm_ls *ls)
	26	{
	27	return test_bit(LSFL_RCOM_READY, &ls->ls_flags);
	28	}
	29
a070a91c AA	30	static void _create_rcom(struct dlm_ls *ls, int to_nodeid, int type, int len,
a070a91c AA	31	struct dlm_rcom *rc_ret, char mb, int mb_len)
e7fd4179 DT	32	{
e7fd4179 DT	33	struct dlm_rcom *rc;
e7fd4179 DT	34
	35	rc = (struct dlm_rcom *) mb;
	36
	37	rc->rc_header.h_version = (DLM_HEADER_MAJOR \| DLM_HEADER_MINOR);
8e2e4086	38	rc->rc_header.u.h_lockspace = ls->ls_global_id;
e7fd4179 DT	39	rc->rc_header.h_nodeid = dlm_our_nodeid();
	40	rc->rc_header.h_length = mb_len;
	41	rc->rc_header.h_cmd = DLM_RCOM;
	42
	43	rc->rc_type = type;
	44
38aa8b0c DT	45	spin_lock(&ls->ls_recover_lock);
	46	rc->rc_seq = ls->ls_recover_seq;
	47	spin_unlock(&ls->ls_recover_lock);
	48
e7fd4179	49	*rc_ret = rc;
a070a91c AA	50	}
	51
	52	static int create_rcom(struct dlm_ls *ls, int to_nodeid, int type, int len,
	53	struct dlm_rcom rc_ret, struct dlm_mhandle mh_ret)
	54	{
	55	int mb_len = sizeof(struct dlm_rcom) + len;
	56	struct dlm_mhandle *mh;
	57	char *mb;
	58
	59	mh = dlm_midcomms_get_mhandle(to_nodeid, mb_len, GFP_NOFS, &mb);
	60	if (!mh) {
	61	log_print("%s to %d type %d len %d ENOBUFS",
	62	__func__, to_nodeid, type, len);
	63	return -ENOBUFS;
	64	}
	65
	66	_create_rcom(ls, to_nodeid, type, len, rc_ret, mb, mb_len);
	67	*mh_ret = mh;
e7fd4179 DT	68	return 0;
	69	}
	70
a070a91c AA	71	static int create_rcom_stateless(struct dlm_ls *ls, int to_nodeid, int type,
a070a91c AA	72	int len, struct dlm_rcom **rc_ret,
8f2dc78d	73	struct dlm_msg **msg_ret)
a070a91c AA	74	{
a070a91c AA	75	int mb_len = sizeof(struct dlm_rcom) + len;
8f2dc78d	76	struct dlm_msg *msg;
a070a91c AA	77	char *mb;
a070a91c AA	78
8f2dc78d AA	79	msg = dlm_lowcomms_new_msg(to_nodeid, mb_len, GFP_NOFS, &mb,
	80	NULL, NULL);
	81	if (!msg) {
a070a91c AA	82	log_print("create_rcom to %d type %d len %d ENOBUFS",
	83	to_nodeid, type, len);
	84	return -ENOBUFS;
	85	}
	86
	87	_create_rcom(ls, to_nodeid, type, len, rc_ret, mb, mb_len);
8f2dc78d	88	*msg_ret = msg;
a070a91c AA	89	return 0;
	90	}
	91
	92	static void _send_rcom(struct dlm_ls ls, struct dlm_rcom rc)
	93	{
	94	dlm_rcom_out(rc);
	95	}
	96
e7fd4179 DT	97	static void send_rcom(struct dlm_ls ls, struct dlm_mhandle mh,
	98	struct dlm_rcom *rc)
	99	{
a070a91c AA	100	_send_rcom(ls, rc);
	101	dlm_midcomms_commit_mhandle(mh);
	102	}
	103
8f2dc78d	104	static void send_rcom_stateless(struct dlm_ls ls, struct dlm_msg msg,
a070a91c AA	105	struct dlm_rcom *rc)
	106	{
	107	_send_rcom(ls, rc);
8f2dc78d AA	108	dlm_lowcomms_commit_msg(msg);
8f2dc78d AA	109	dlm_lowcomms_put_msg(msg);
e7fd4179 DT	110	}
e7fd4179 DT	111
757a4271 DT	112	static void set_rcom_status(struct dlm_ls ls, struct rcom_status rs,
	113	uint32_t flags)
	114	{
	115	rs->rs_flags = cpu_to_le32(flags);
	116	}
	117
e7fd4179 DT	118	/* When replying to a status request, a node also sends back its
	119	configuration values. The requesting node then checks that the remote
	120	node is configured the same way as itself. */
	121
757a4271 DT	122	static void set_rcom_config(struct dlm_ls ls, struct rcom_config rf,
757a4271 DT	123	uint32_t num_slots)
e7fd4179	124	{
93ff2971 AV	125	rf->rf_lvblen = cpu_to_le32(ls->ls_lvblen);
93ff2971 AV	126	rf->rf_lsflags = cpu_to_le32(ls->ls_exflags);
757a4271 DT	127
	128	rf->rf_our_slot = cpu_to_le16(ls->ls_slot);
	129	rf->rf_num_slots = cpu_to_le16(num_slots);
	130	rf->rf_generation = cpu_to_le32(ls->ls_generation);
e7fd4179 DT	131	}
e7fd4179 DT	132
757a4271	133	static int check_rcom_config(struct dlm_ls ls, struct dlm_rcom rc, int nodeid)
e7fd4179	134	{
9e971b71 DT	135	struct rcom_config rf = (struct rcom_config ) rc->rc_buf;
	136
	137	if ((rc->rc_header.h_version & 0xFFFF0000) != DLM_HEADER_MAJOR) {
	138	log_error(ls, "version mismatch: %x nodeid %d: %x",
	139	DLM_HEADER_MAJOR \| DLM_HEADER_MINOR, nodeid,
	140	rc->rc_header.h_version);
8b0e7b2c	141	return -EPROTO;
9e971b71 DT	142	}
9e971b71 DT	143
93ff2971 AV	144	if (le32_to_cpu(rf->rf_lvblen) != ls->ls_lvblen \|\|
93ff2971 AV	145	le32_to_cpu(rf->rf_lsflags) != ls->ls_exflags) {
e7fd4179	146	log_error(ls, "config mismatch: %d,%x nodeid %d: %d,%x",
93ff2971 AV	147	ls->ls_lvblen, ls->ls_exflags, nodeid,
	148	le32_to_cpu(rf->rf_lvblen),
	149	le32_to_cpu(rf->rf_lsflags));
8b0e7b2c	150	return -EPROTO;
e7fd4179 DT	151	}
	152	return 0;
	153	}
	154
98f176fb DT	155	static void allow_sync_reply(struct dlm_ls ls, uint64_t new_seq)
	156	{
	157	spin_lock(&ls->ls_rcom_spin);
	158	*new_seq = ++ls->ls_rcom_seq;
	159	set_bit(LSFL_RCOM_WAIT, &ls->ls_flags);
	160	spin_unlock(&ls->ls_rcom_spin);
	161	}
	162
	163	static void disallow_sync_reply(struct dlm_ls *ls)
	164	{
	165	spin_lock(&ls->ls_rcom_spin);
	166	clear_bit(LSFL_RCOM_WAIT, &ls->ls_flags);
	167	clear_bit(LSFL_RCOM_READY, &ls->ls_flags);
	168	spin_unlock(&ls->ls_rcom_spin);
	169	}
	170
757a4271 DT	171	/*
	172	* low nodeid gathers one slot value at a time from each node.
	173	* it sets need_slots=0, and saves rf_our_slot returned from each
	174	* rcom_config.
	175	*
	176	* other nodes gather all slot values at once from the low nodeid.
	177	* they set need_slots=1, and ignore the rf_our_slot returned from each
	178	* rcom_config. they use the rf_num_slots returned from the low
	179	* node's rcom_config.
	180	*/
	181
	182	int dlm_rcom_status(struct dlm_ls *ls, int nodeid, uint32_t status_flags)
e7fd4179 DT	183	{
e7fd4179 DT	184	struct dlm_rcom *rc;
8f2dc78d	185	struct dlm_msg *msg;
e7fd4179 DT	186	int error = 0;
e7fd4179 DT	187
faa0f267	188	ls->ls_recover_nodeid = nodeid;
e7fd4179 DT	189
e7fd4179 DT	190	if (nodeid == dlm_our_nodeid()) {
4007685c	191	rc = ls->ls_recover_buf;
e7fd4179 DT	192	rc->rc_result = dlm_recover_status(ls);
	193	goto out;
	194	}
	195
59661212	196	retry:
a070a91c	197	error = create_rcom_stateless(ls, nodeid, DLM_RCOM_STATUS,
8f2dc78d	198	sizeof(struct rcom_status), &rc, &msg);
e7fd4179 DT	199	if (error)
e7fd4179 DT	200	goto out;
98f176fb	201
757a4271 DT	202	set_rcom_status(ls, (struct rcom_status *)rc->rc_buf, status_flags);
757a4271 DT	203
98f176fb	204	allow_sync_reply(ls, &rc->rc_id);
489d8e55	205	memset(ls->ls_recover_buf, 0, DEFAULT_BUFFER_SIZE);
e7fd4179	206
8f2dc78d	207	send_rcom_stateless(ls, msg, rc);
e7fd4179 DT	208
e7fd4179 DT	209	error = dlm_wait_function(ls, &rcom_response);
98f176fb	210	disallow_sync_reply(ls);
59661212	211	if (error == -ETIMEDOUT)
59661212	212	goto retry;
e7fd4179 DT	213	if (error)
	214	goto out;
	215
4007685c	216	rc = ls->ls_recover_buf;
e7fd4179 DT	217
	218	if (rc->rc_result == -ESRCH) {
	219	/* we pretend the remote lockspace exists with 0 status */
	220	log_debug(ls, "remote node %d not ready", nodeid);
	221	rc->rc_result = 0;
757a4271 DT	222	error = 0;
	223	} else {
	224	error = check_rcom_config(ls, rc, nodeid);
	225	}
	226
e7fd4179 DT	227	/* the caller looks at rc_result for the remote recovery status */
	228	out:
	229	return error;
	230	}
	231
	232	static void receive_rcom_status(struct dlm_ls ls, struct dlm_rcom rc_in)
	233	{
	234	struct dlm_rcom *rc;
757a4271 DT	235	struct rcom_status *rs;
	236	uint32_t status;
	237	int nodeid = rc_in->rc_header.h_nodeid;
	238	int len = sizeof(struct rcom_config);
8f2dc78d	239	struct dlm_msg *msg;
757a4271 DT	240	int num_slots = 0;
	241	int error;
	242
	243	if (!dlm_slots_version(&rc_in->rc_header)) {
	244	status = dlm_recover_status(ls);
	245	goto do_create;
	246	}
	247
	248	rs = (struct rcom_status *)rc_in->rc_buf;
e7fd4179	249
c07127b4	250	if (!(le32_to_cpu(rs->rs_flags) & DLM_RSF_NEED_SLOTS)) {
757a4271 DT	251	status = dlm_recover_status(ls);
	252	goto do_create;
	253	}
	254
	255	spin_lock(&ls->ls_recover_lock);
	256	status = ls->ls_recover_status;
	257	num_slots = ls->ls_num_slots;
	258	spin_unlock(&ls->ls_recover_lock);
	259	len += num_slots * sizeof(struct rcom_slot);
	260
	261	do_create:
a070a91c	262	error = create_rcom_stateless(ls, nodeid, DLM_RCOM_STATUS_REPLY,
8f2dc78d	263	len, &rc, &msg);
e7fd4179 DT	264	if (error)
e7fd4179 DT	265	return;
757a4271	266
4a99c3d9	267	rc->rc_id = rc_in->rc_id;
38aa8b0c	268	rc->rc_seq_reply = rc_in->rc_seq;
757a4271 DT	269	rc->rc_result = status;
	270
	271	set_rcom_config(ls, (struct rcom_config *)rc->rc_buf, num_slots);
	272
	273	if (!num_slots)
	274	goto do_send;
	275
	276	spin_lock(&ls->ls_recover_lock);
	277	if (ls->ls_num_slots != num_slots) {
	278	spin_unlock(&ls->ls_recover_lock);
	279	log_debug(ls, "receive_rcom_status num_slots %d to %d",
	280	num_slots, ls->ls_num_slots);
	281	rc->rc_result = 0;
	282	set_rcom_config(ls, (struct rcom_config *)rc->rc_buf, 0);
	283	goto do_send;
	284	}
	285
	286	dlm_slots_copy_out(ls, rc);
	287	spin_unlock(&ls->ls_recover_lock);
e7fd4179	288
757a4271	289	do_send:
8f2dc78d	290	send_rcom_stateless(ls, msg, rc);
e7fd4179 DT	291	}
e7fd4179 DT	292
4a99c3d9	293	static void receive_sync_reply(struct dlm_ls ls, struct dlm_rcom rc_in)
e7fd4179	294	{
98f176fb DT	295	spin_lock(&ls->ls_rcom_spin);
	296	if (!test_bit(LSFL_RCOM_WAIT, &ls->ls_flags) \|\|
	297	rc_in->rc_id != ls->ls_rcom_seq) {
	298	log_debug(ls, "reject reply %d from %d seq %llx expect %llx",
	299	rc_in->rc_type, rc_in->rc_header.h_nodeid,
57adf7ee RK	300	(unsigned long long)rc_in->rc_id,
57adf7ee RK	301	(unsigned long long)ls->ls_rcom_seq);
98f176fb	302	goto out;
4a99c3d9	303	}
e7fd4179 DT	304	memcpy(ls->ls_recover_buf, rc_in, rc_in->rc_header.h_length);
e7fd4179 DT	305	set_bit(LSFL_RCOM_READY, &ls->ls_flags);
98f176fb	306	clear_bit(LSFL_RCOM_WAIT, &ls->ls_flags);
e7fd4179	307	wake_up(&ls->ls_wait_general);
98f176fb DT	308	out:
98f176fb DT	309	spin_unlock(&ls->ls_rcom_spin);
e7fd4179 DT	310	}
	311
	312	int dlm_rcom_names(struct dlm_ls ls, int nodeid, char last_name, int last_len)
	313	{
	314	struct dlm_rcom *rc;
8f2dc78d	315	struct dlm_msg *msg;
4007685c	316	int error = 0;
e7fd4179	317
faa0f267	318	ls->ls_recover_nodeid = nodeid;
e7fd4179	319
59661212	320	retry:
a070a91c	321	error = create_rcom_stateless(ls, nodeid, DLM_RCOM_NAMES, last_len,
8f2dc78d	322	&rc, &msg);
e7fd4179 DT	323	if (error)
	324	goto out;
	325	memcpy(rc->rc_buf, last_name, last_len);
98f176fb DT	326
98f176fb DT	327	allow_sync_reply(ls, &rc->rc_id);
489d8e55	328	memset(ls->ls_recover_buf, 0, DEFAULT_BUFFER_SIZE);
e7fd4179	329
8f2dc78d	330	send_rcom_stateless(ls, msg, rc);
e7fd4179 DT	331
e7fd4179 DT	332	error = dlm_wait_function(ls, &rcom_response);
98f176fb	333	disallow_sync_reply(ls);
59661212	334	if (error == -ETIMEDOUT)
59661212	335	goto retry;
e7fd4179 DT	336	out:
	337	return error;
	338	}
	339
	340	static void receive_rcom_names(struct dlm_ls ls, struct dlm_rcom rc_in)
	341	{
	342	struct dlm_rcom *rc;
38aa8b0c	343	int error, inlen, outlen, nodeid;
8f2dc78d	344	struct dlm_msg *msg;
e7fd4179 DT	345
	346	nodeid = rc_in->rc_header.h_nodeid;
	347	inlen = rc_in->rc_header.h_length - sizeof(struct dlm_rcom);
9f8f9c77	348	outlen = LOWCOMMS_MAX_TX_BUFFER_LEN - sizeof(struct dlm_rcom);
e7fd4179	349
a070a91c	350	error = create_rcom_stateless(ls, nodeid, DLM_RCOM_NAMES_REPLY, outlen,
8f2dc78d	351	&rc, &msg);
e7fd4179 DT	352	if (error)
e7fd4179 DT	353	return;
4a99c3d9	354	rc->rc_id = rc_in->rc_id;
38aa8b0c	355	rc->rc_seq_reply = rc_in->rc_seq;
e7fd4179 DT	356
	357	dlm_copy_master_names(ls, rc_in->rc_buf, inlen, rc->rc_buf, outlen,
	358	nodeid);
8f2dc78d	359	send_rcom_stateless(ls, msg, rc);
e7fd4179 DT	360	}
e7fd4179 DT	361
e7fd4179 DT	362	int dlm_send_rcom_lookup(struct dlm_rsb *r, int dir_nodeid)
	363	{
	364	struct dlm_rcom *rc;
	365	struct dlm_mhandle *mh;
	366	struct dlm_ls *ls = r->res_ls;
	367	int error;
	368
	369	error = create_rcom(ls, dir_nodeid, DLM_RCOM_LOOKUP, r->res_length,
	370	&rc, &mh);
	371	if (error)
	372	goto out;
	373	memcpy(rc->rc_buf, r->res_name, r->res_length);
1d7c484e	374	rc->rc_id = (unsigned long) r->res_id;
e7fd4179 DT	375
	376	send_rcom(ls, mh, rc);
	377	out:
	378	return error;
	379	}
	380
	381	static void receive_rcom_lookup(struct dlm_ls ls, struct dlm_rcom rc_in)
	382	{
	383	struct dlm_rcom *rc;
	384	struct dlm_mhandle *mh;
	385	int error, ret_nodeid, nodeid = rc_in->rc_header.h_nodeid;
	386	int len = rc_in->rc_header.h_length - sizeof(struct dlm_rcom);
	387
9250e523	388	/* Old code would send this special id to trigger a debug dump. */
c04fecb4 DT	389	if (rc_in->rc_id == 0xFFFFFFFF) {
	390	log_error(ls, "receive_rcom_lookup dump from %d", nodeid);
	391	dlm_dump_rsb_name(ls, rc_in->rc_buf, len);
	392	return;
	393	}
	394
f6089981 CIK	395	error = create_rcom(ls, nodeid, DLM_RCOM_LOOKUP_REPLY, 0, &rc, &mh);
	396	if (error)
	397	return;
	398
c04fecb4 DT	399	error = dlm_master_lookup(ls, nodeid, rc_in->rc_buf, len,
c04fecb4 DT	400	DLM_LU_RECOVER_MASTER, &ret_nodeid, NULL);
e7fd4179 DT	401	if (error)
	402	ret_nodeid = error;
	403	rc->rc_result = ret_nodeid;
	404	rc->rc_id = rc_in->rc_id;
38aa8b0c	405	rc->rc_seq_reply = rc_in->rc_seq;
e7fd4179 DT	406
	407	send_rcom(ls, mh, rc);
	408	}
	409
	410	static void receive_rcom_lookup_reply(struct dlm_ls ls, struct dlm_rcom rc_in)
	411	{
	412	dlm_recover_master_reply(ls, rc_in);
	413	}
	414
	415	static void pack_rcom_lock(struct dlm_rsb r, struct dlm_lkb lkb,
	416	struct rcom_lock *rl)
	417	{
	418	memset(rl, 0, sizeof(*rl));
	419
163a1859 AV	420	rl->rl_ownpid = cpu_to_le32(lkb->lkb_ownpid);
	421	rl->rl_lkid = cpu_to_le32(lkb->lkb_id);
	422	rl->rl_exflags = cpu_to_le32(lkb->lkb_exflags);
	423	rl->rl_flags = cpu_to_le32(lkb->lkb_flags);
	424	rl->rl_lvbseq = cpu_to_le32(lkb->lkb_lvbseq);
e7fd4179 DT	425	rl->rl_rqmode = lkb->lkb_rqmode;
	426	rl->rl_grmode = lkb->lkb_grmode;
	427	rl->rl_status = lkb->lkb_status;
163a1859	428	rl->rl_wait_type = cpu_to_le16(lkb->lkb_wait_type);
e7fd4179	429
e5dae548	430	if (lkb->lkb_bastfn)
8304d6f2	431	rl->rl_asts \|= DLM_CB_BAST;
e5dae548	432	if (lkb->lkb_astfn)
8304d6f2	433	rl->rl_asts \|= DLM_CB_CAST;
e7fd4179	434
163a1859	435	rl->rl_namelen = cpu_to_le16(r->res_length);
e7fd4179 DT	436	memcpy(rl->rl_name, r->res_name, r->res_length);
	437
	438	/* FIXME: might we have an lvb without DLM_LKF_VALBLK set ?
	439	If so, receive_rcom_lock_args() won't take this copy. */
	440
	441	if (lkb->lkb_lvbptr)
	442	memcpy(rl->rl_lvb, lkb->lkb_lvbptr, r->res_ls->ls_lvblen);
	443	}
	444
	445	int dlm_send_rcom_lock(struct dlm_rsb r, struct dlm_lkb lkb)
	446	{
	447	struct dlm_ls *ls = r->res_ls;
	448	struct dlm_rcom *rc;
	449	struct dlm_mhandle *mh;
	450	struct rcom_lock *rl;
	451	int error, len = sizeof(struct rcom_lock);
	452
	453	if (lkb->lkb_lvbptr)
	454	len += ls->ls_lvblen;
	455
	456	error = create_rcom(ls, r->res_nodeid, DLM_RCOM_LOCK, len, &rc, &mh);
	457	if (error)
	458	goto out;
	459
	460	rl = (struct rcom_lock *) rc->rc_buf;
	461	pack_rcom_lock(r, lkb, rl);
	462	rc->rc_id = (unsigned long) r;
	463
	464	send_rcom(ls, mh, rc);
	465	out:
	466	return error;
	467	}
	468
ae773d0b	469	/* needs at least dlm_rcom + rcom_lock */
e7fd4179 DT	470	static void receive_rcom_lock(struct dlm_ls ls, struct dlm_rcom rc_in)
	471	{
	472	struct dlm_rcom *rc;
	473	struct dlm_mhandle *mh;
	474	int error, nodeid = rc_in->rc_header.h_nodeid;
	475
	476	dlm_recover_master_copy(ls, rc_in);
	477
	478	error = create_rcom(ls, nodeid, DLM_RCOM_LOCK_REPLY,
	479	sizeof(struct rcom_lock), &rc, &mh);
	480	if (error)
	481	return;
	482
	483	/* We send back the same rcom_lock struct we received, but
	484	dlm_recover_master_copy() has filled in rl_remid and rl_result */
	485
	486	memcpy(rc->rc_buf, rc_in->rc_buf, sizeof(struct rcom_lock));
	487	rc->rc_id = rc_in->rc_id;
38aa8b0c	488	rc->rc_seq_reply = rc_in->rc_seq;
e7fd4179 DT	489
	490	send_rcom(ls, mh, rc);
	491	}
	492
c36258b5 DT	493	/* If the lockspace doesn't exist then still send a status message
	494	back; it's possible that it just doesn't have its global_id yet. */
	495
	496	int dlm_send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in)
e7fd4179 DT	497	{
e7fd4179 DT	498	struct dlm_rcom *rc;
1babdb45	499	struct rcom_config *rf;
e7fd4179 DT	500	struct dlm_mhandle *mh;
e7fd4179 DT	501	char *mb;
1babdb45	502	int mb_len = sizeof(struct dlm_rcom) + sizeof(struct rcom_config);
e7fd4179	503
a070a91c	504	mh = dlm_midcomms_get_mhandle(nodeid, mb_len, GFP_NOFS, &mb);
e7fd4179 DT	505	if (!mh)
e7fd4179 DT	506	return -ENOBUFS;
e7fd4179 DT	507
	508	rc = (struct dlm_rcom *) mb;
	509
	510	rc->rc_header.h_version = (DLM_HEADER_MAJOR \| DLM_HEADER_MINOR);
8e2e4086	511	rc->rc_header.u.h_lockspace = rc_in->rc_header.u.h_lockspace;
e7fd4179 DT	512	rc->rc_header.h_nodeid = dlm_our_nodeid();
	513	rc->rc_header.h_length = mb_len;
	514	rc->rc_header.h_cmd = DLM_RCOM;
	515
	516	rc->rc_type = DLM_RCOM_STATUS_REPLY;
f5888750	517	rc->rc_id = rc_in->rc_id;
38aa8b0c	518	rc->rc_seq_reply = rc_in->rc_seq;
e7fd4179 DT	519	rc->rc_result = -ESRCH;
e7fd4179 DT	520
1babdb45	521	rf = (struct rcom_config *) rc->rc_buf;
93ff2971	522	rf->rf_lvblen = cpu_to_le32(~0U);
1babdb45	523
e7fd4179	524	dlm_rcom_out(rc);
a070a91c	525	dlm_midcomms_commit_mhandle(mh);
e7fd4179 DT	526
	527	return 0;
	528	}
	529
c04fecb4 DT	530	/*
	531	* Ignore messages for stage Y before we set
	532	* recover_status bit for stage X:
	533	*
	534	* recover_status = 0
	535	*
	536	* dlm_recover_members()
	537	* - send nothing
	538	* - recv nothing
	539	* - ignore NAMES, NAMES_REPLY
	540	* - ignore LOOKUP, LOOKUP_REPLY
	541	* - ignore LOCK, LOCK_REPLY
	542	*
	543	* recover_status \|= NODES
	544	*
	545	* dlm_recover_members_wait()
	546	*
	547	* dlm_recover_directory()
	548	* - send NAMES
	549	* - recv NAMES_REPLY
	550	* - ignore LOOKUP, LOOKUP_REPLY
	551	* - ignore LOCK, LOCK_REPLY
	552	*
	553	* recover_status \|= DIR
	554	*
	555	* dlm_recover_directory_wait()
	556	*
	557	* dlm_recover_masters()
	558	* - send LOOKUP
	559	* - recv LOOKUP_REPLY
	560	*
	561	* dlm_recover_locks()
	562	* - send LOCKS
	563	* - recv LOCKS_REPLY
	564	*
	565	* recover_status \|= LOCKS
	566	*
	567	* dlm_recover_locks_wait()
	568	*
	569	* recover_status \|= DONE
	570	*/
	571
d6e24788 DT	572	/* Called by dlm_recv; corresponds to dlm_receive_message() but special
	573	recovery-only comms are sent through here. */
	574
	575	void dlm_receive_rcom(struct dlm_ls ls, struct dlm_rcom rc, int nodeid)
38aa8b0c	576	{
d6e24788	577	int lock_size = sizeof(struct dlm_rcom) + sizeof(struct rcom_lock);
c04fecb4	578	int stop, reply = 0, names = 0, lookup = 0, lock = 0;
4875647a	579	uint32_t status;
38aa8b0c	580	uint64_t seq;
38aa8b0c DT	581
38aa8b0c DT	582	switch (rc->rc_type) {
c04fecb4 DT	583	case DLM_RCOM_STATUS_REPLY:
	584	reply = 1;
	585	break;
	586	case DLM_RCOM_NAMES:
	587	names = 1;
	588	break;
	589	case DLM_RCOM_NAMES_REPLY:
	590	names = 1;
	591	reply = 1;
	592	break;
	593	case DLM_RCOM_LOOKUP:
	594	lookup = 1;
	595	break;
	596	case DLM_RCOM_LOOKUP_REPLY:
	597	lookup = 1;
	598	reply = 1;
	599	break;
4875647a DT	600	case DLM_RCOM_LOCK:
	601	lock = 1;
	602	break;
	603	case DLM_RCOM_LOCK_REPLY:
	604	lock = 1;
	605	reply = 1;
	606	break;
90db4f8b	607	}
38aa8b0c	608
d6e24788	609	spin_lock(&ls->ls_recover_lock);
4875647a	610	status = ls->ls_recover_status;
475f230c	611	stop = test_bit(LSFL_RECOVER_STOP, &ls->ls_flags);
d6e24788 DT	612	seq = ls->ls_recover_seq;
d6e24788 DT	613	spin_unlock(&ls->ls_recover_lock);
ae773d0b	614
c04fecb4 DT	615	if (stop && (rc->rc_type != DLM_RCOM_STATUS))
	616	goto ignore;
	617
	618	if (reply && (rc->rc_seq_reply != seq))
	619	goto ignore;
	620
	621	if (!(status & DLM_RS_NODES) && (names \|\| lookup \|\| lock))
	622	goto ignore;
	623
	624	if (!(status & DLM_RS_DIR) && (lookup \|\| lock))
	625	goto ignore;
e7fd4179	626
e7fd4179 DT	627	switch (rc->rc_type) {
	628	case DLM_RCOM_STATUS:
	629	receive_rcom_status(ls, rc);
	630	break;
	631
	632	case DLM_RCOM_NAMES:
	633	receive_rcom_names(ls, rc);
	634	break;
	635
	636	case DLM_RCOM_LOOKUP:
	637	receive_rcom_lookup(ls, rc);
	638	break;
	639
	640	case DLM_RCOM_LOCK:
ae773d0b AV	641	if (rc->rc_header.h_length < lock_size)
ae773d0b AV	642	goto Eshort;
e7fd4179 DT	643	receive_rcom_lock(ls, rc);
	644	break;
	645
	646	case DLM_RCOM_STATUS_REPLY:
dbcfc347	647	receive_sync_reply(ls, rc);
e7fd4179 DT	648	break;
	649
	650	case DLM_RCOM_NAMES_REPLY:
dbcfc347	651	receive_sync_reply(ls, rc);
e7fd4179 DT	652	break;
	653
	654	case DLM_RCOM_LOOKUP_REPLY:
	655	receive_rcom_lookup_reply(ls, rc);
	656	break;
	657
	658	case DLM_RCOM_LOCK_REPLY:
ae773d0b AV	659	if (rc->rc_header.h_length < lock_size)
ae773d0b AV	660	goto Eshort;
dbcfc347	661	dlm_recover_process_copy(ls, rc);
e7fd4179 DT	662	break;
	663
	664	default:
dbcfc347	665	log_error(ls, "receive_rcom bad type %d", rc->rc_type);
e7fd4179	666	}
c04fecb4 DT	667	return;
	668
	669	ignore:
	670	log_limit(ls, "dlm_receive_rcom ignore msg %d "
	671	"from %d %llu %llu recover seq %llu sts %x gen %u",
	672	rc->rc_type,
	673	nodeid,
	674	(unsigned long long)rc->rc_seq,
	675	(unsigned long long)rc->rc_seq_reply,
	676	(unsigned long long)seq,
	677	status, ls->ls_generation);
c36258b5	678	return;
ae773d0b	679	Eshort:
c04fecb4 DT	680	log_error(ls, "recovery message %d from %d is too short",
c04fecb4 DT	681	rc->rc_type, nodeid);
e7fd4179 DT	682	}
e7fd4179 DT	683