[deliverable/linux.git] / fs / nfs / callback_proc.c

/*
 * linux/fs/nfs/callback_proc.c
 *
 * Copyright (C) 2004 Trond Myklebust
 *
 * NFSv4 callback procedures
 */
#include <linux/nfs4.h>
#include <linux/nfs_fs.h>
#include <linux/slab.h>
#include <linux/rcupdate.h>
#include "nfs4_fs.h"
#include "callback.h"
#include "delegation.h"
#include "internal.h"
#include "pnfs.h"
#include "nfs4session.h"
#include "nfs4trace.h"

#define NFSDBG_FACILITY NFSDBG_CALLBACK

__be32 nfs4_callback_getattr(struct cb_getattrargs *args,
			     struct cb_getattrres *res,
			     struct cb_process_state *cps)
{
	struct nfs_delegation *delegation;
	struct nfs_inode *nfsi;
	struct inode *inode;

	res->status = htonl(NFS4ERR_OP_NOT_IN_SESSION);
	if (!cps->clp) /* Always set for v4.0. Set in cb_sequence for v4.1 */
		goto out;

	res->bitmap[0] = res->bitmap[1] = 0;
	res->status = htonl(NFS4ERR_BADHANDLE);

	dprintk_rcu("NFS: GETATTR callback request from %s\n",
		rpc_peeraddr2str(cps->clp->cl_rpcclient, RPC_DISPLAY_ADDR));

	inode = nfs_delegation_find_inode(cps->clp, &args->fh);
	if (inode == NULL) {
		trace_nfs4_cb_getattr(cps->clp, &args->fh, NULL,
				-ntohl(res->status));
		goto out;
	}
	nfsi = NFS_I(inode);
	rcu_read_lock();
	delegation = rcu_dereference(nfsi->delegation);
	if (delegation == NULL || (delegation->type & FMODE_WRITE) == 0)
		goto out_iput;
	res->size = i_size_read(inode);
	res->change_attr = delegation->change_attr;
	if (nfsi->nrequests != 0)
		res->change_attr++;
	res->ctime = inode->i_ctime;
	res->mtime = inode->i_mtime;
	res->bitmap[0] = (FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE) &
		args->bitmap[0];
	res->bitmap[1] = (FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY) &
		args->bitmap[1];
	res->status = 0;
out_iput:
	rcu_read_unlock();
	trace_nfs4_cb_getattr(cps->clp, &args->fh, inode, -ntohl(res->status));
	iput(inode);
out:
	dprintk("%s: exit with status = %d\n", __func__, ntohl(res->status));
	return res->status;
}

__be32 nfs4_callback_recall(struct cb_recallargs *args, void *dummy,
			    struct cb_process_state *cps)
{
	struct inode *inode;
	__be32 res;
	
	res = htonl(NFS4ERR_OP_NOT_IN_SESSION);
	if (!cps->clp) /* Always set for v4.0. Set in cb_sequence for v4.1 */
		goto out;

	dprintk_rcu("NFS: RECALL callback request from %s\n",
		rpc_peeraddr2str(cps->clp->cl_rpcclient, RPC_DISPLAY_ADDR));

	res = htonl(NFS4ERR_BADHANDLE);
	inode = nfs_delegation_find_inode(cps->clp, &args->fh);
	if (inode == NULL)
		goto out;
	/* Set up a helper thread to actually return the delegation */
	switch (nfs_async_inode_return_delegation(inode, &args->stateid)) {
	case 0:
		res = 0;
		break;
	case -ENOENT:
		res = htonl(NFS4ERR_BAD_STATEID);
		break;
	default:
		res = htonl(NFS4ERR_RESOURCE);
	}
	trace_nfs4_recall_delegation(inode, -ntohl(res));
	iput(inode);
out:
	dprintk("%s: exit with status = %d\n", __func__, ntohl(res));
	return res;
}

#if defined(CONFIG_NFS_V4_1)

/*
 * Lookup a layout by filehandle.
 *
 * Note: gets a refcount on the layout hdr and on its respective inode.
 * Caller must put the layout hdr and the inode.
 *
 * TODO: keep track of all layouts (and delegations) in a hash table
 * hashed by filehandle.
 */
static struct pnfs_layout_hdr * get_layout_by_fh_locked(struct nfs_client *clp,
		struct nfs_fh *fh, nfs4_stateid *stateid)
{
	struct nfs_server *server;
	struct inode *ino;
	struct pnfs_layout_hdr *lo;

	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
		list_for_each_entry(lo, &server->layouts, plh_layouts) {
			if (!nfs4_stateid_match_other(&lo->plh_stateid, stateid))
				continue;
			if (nfs_compare_fh(fh, &NFS_I(lo->plh_inode)->fh))
				continue;
			ino = igrab(lo->plh_inode);
			if (!ino)
				break;
			spin_lock(&ino->i_lock);
			/* Is this layout in the process of being freed? */
			if (NFS_I(ino)->layout != lo) {
				spin_unlock(&ino->i_lock);
				iput(ino);
				break;
			}
			pnfs_get_layout_hdr(lo);
			spin_unlock(&ino->i_lock);
			return lo;
		}
	}

	return NULL;
}

static struct pnfs_layout_hdr * get_layout_by_fh(struct nfs_client *clp,
		struct nfs_fh *fh, nfs4_stateid *stateid)
{
	struct pnfs_layout_hdr *lo;

	spin_lock(&clp->cl_lock);
	rcu_read_lock();
	lo = get_layout_by_fh_locked(clp, fh, stateid);
	rcu_read_unlock();
	spin_unlock(&clp->cl_lock);

	return lo;
}

/*
 * Enforce RFC5661 section 12.5.5.2.1. (Layout Recall and Return Sequencing)
 */
static bool pnfs_check_stateid_sequence(struct pnfs_layout_hdr *lo,
					const nfs4_stateid *new)
{
	u32 oldseq, newseq;

	oldseq = be32_to_cpu(lo->plh_stateid.seqid);
	newseq = be32_to_cpu(new->seqid);

	if (newseq > oldseq + 1)
		return false;
	return true;
}

static u32 initiate_file_draining(struct nfs_client *clp,
				  struct cb_layoutrecallargs *args)
{
	struct inode *ino;
	struct pnfs_layout_hdr *lo;
	u32 rv = NFS4ERR_NOMATCHING_LAYOUT;
	LIST_HEAD(free_me_list);

	lo = get_layout_by_fh(clp, &args->cbl_fh, &args->cbl_stateid);
	if (!lo)
		goto out;

	ino = lo->plh_inode;

	spin_lock(&ino->i_lock);
	if (!pnfs_check_stateid_sequence(lo, &args->cbl_stateid)) {
		rv = NFS4ERR_DELAY;
		goto unlock;
	}
	pnfs_set_layout_stateid(lo, &args->cbl_stateid, true);
	spin_unlock(&ino->i_lock);

	pnfs_layoutcommit_inode(ino, false);

	spin_lock(&ino->i_lock);
	/*
	 * Enforce RFC5661 Section 12.5.5.2.1.5 (Bulk Recall and Return)
	 */
	if (test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags)) {
		rv = NFS4ERR_DELAY;
		goto unlock;
	}

	if (pnfs_mark_matching_lsegs_invalid(lo, &free_me_list,
					&args->cbl_range)) {
		pnfs_mark_matching_lsegs_return(lo,
				&free_me_list,
				&args->cbl_range);
		rv = NFS4ERR_DELAY;
		goto unlock;
	}

	if (NFS_SERVER(ino)->pnfs_curr_ld->return_range) {
		NFS_SERVER(ino)->pnfs_curr_ld->return_range(lo,
			&args->cbl_range);
	}
	pnfs_mark_layout_returned_if_empty(lo);
unlock:
	spin_unlock(&ino->i_lock);
	pnfs_free_lseg_list(&free_me_list);
	pnfs_put_layout_hdr(lo);
	trace_nfs4_cb_layoutrecall_inode(clp, &args->cbl_fh, ino, -rv);
	iput(ino);
out:
	return rv;
}

static u32 initiate_bulk_draining(struct nfs_client *clp,
				  struct cb_layoutrecallargs *args)
{
	int stat;

	if (args->cbl_recall_type == RETURN_FSID)
		stat = pnfs_destroy_layouts_byfsid(clp, &args->cbl_fsid, true);
	else
		stat = pnfs_destroy_layouts_byclid(clp, true);
	if (stat != 0)
		return NFS4ERR_DELAY;
	return NFS4ERR_NOMATCHING_LAYOUT;
}

static u32 do_callback_layoutrecall(struct nfs_client *clp,
				    struct cb_layoutrecallargs *args)
{
	u32 res;

	dprintk("%s enter, type=%i\n", __func__, args->cbl_recall_type);
	if (args->cbl_recall_type == RETURN_FILE)
		res = initiate_file_draining(clp, args);
	else
		res = initiate_bulk_draining(clp, args);
	dprintk("%s returning %i\n", __func__, res);
	return res;

}

__be32 nfs4_callback_layoutrecall(struct cb_layoutrecallargs *args,
				  void *dummy, struct cb_process_state *cps)
{
	u32 res;

	dprintk("%s: -->\n", __func__);

	if (cps->clp)
		res = do_callback_layoutrecall(cps->clp, args);
	else
		res = NFS4ERR_OP_NOT_IN_SESSION;

	dprintk("%s: exit with status = %d\n", __func__, res);
	return cpu_to_be32(res);
}

static void pnfs_recall_all_layouts(struct nfs_client *clp)
{
	struct cb_layoutrecallargs args;

	/* Pretend we got a CB_LAYOUTRECALL(ALL) */
	memset(&args, 0, sizeof(args));
	args.cbl_recall_type = RETURN_ALL;
	/* FIXME we ignore errors, what should we do? */
	do_callback_layoutrecall(clp, &args);
}

__be32 nfs4_callback_devicenotify(struct cb_devicenotifyargs *args,
				  void *dummy, struct cb_process_state *cps)
{
	int i;
	__be32 res = 0;
	struct nfs_client *clp = cps->clp;
	struct nfs_server *server = NULL;

	dprintk("%s: -->\n", __func__);

	if (!clp) {
		res = cpu_to_be32(NFS4ERR_OP_NOT_IN_SESSION);
		goto out;
	}

	for (i = 0; i < args->ndevs; i++) {
		struct cb_devicenotifyitem *dev = &args->devs[i];

		if (!server ||
		    server->pnfs_curr_ld->id != dev->cbd_layout_type) {
			rcu_read_lock();
			list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
				if (server->pnfs_curr_ld &&
				    server->pnfs_curr_ld->id == dev->cbd_layout_type) {
					rcu_read_unlock();
					goto found;
				}
			rcu_read_unlock();
			dprintk("%s: layout type %u not found\n",
				__func__, dev->cbd_layout_type);
			continue;
		}

	found:
		nfs4_delete_deviceid(server->pnfs_curr_ld, clp, &dev->cbd_dev_id);
	}

out:
	kfree(args->devs);
	dprintk("%s: exit with status = %u\n",
		__func__, be32_to_cpu(res));
	return res;
}

/*
 * Validate the sequenceID sent by the server.
 * Return success if the sequenceID is one more than what we last saw on
 * this slot, accounting for wraparound.  Increments the slot's sequence.
 *
 * We don't yet implement a duplicate request cache, instead we set the
 * back channel ca_maxresponsesize_cached to zero. This is OK for now
 * since we only currently implement idempotent callbacks anyway.
 *
 * We have a single slot backchannel at this time, so we don't bother
 * checking the used_slots bit array on the table.  The lower layer guarantees
 * a single outstanding callback request at a time.
 */
static __be32
validate_seqid(struct nfs4_slot_table *tbl, struct cb_sequenceargs * args)
{
	struct nfs4_slot *slot;

	dprintk("%s enter. slotid %u seqid %u\n",
		__func__, args->csa_slotid, args->csa_sequenceid);

	if (args->csa_slotid >= NFS41_BC_MAX_CALLBACKS)
		return htonl(NFS4ERR_BADSLOT);

	slot = tbl->slots + args->csa_slotid;
	dprintk("%s slot table seqid: %u\n", __func__, slot->seq_nr);

	/* Normal */
	if (likely(args->csa_sequenceid == slot->seq_nr + 1))
		goto out_ok;

	/* Replay */
	if (args->csa_sequenceid == slot->seq_nr) {
		dprintk("%s seqid %u is a replay\n",
			__func__, args->csa_sequenceid);
		/* Signal process_op to set this error on next op */
		if (args->csa_cachethis == 0)
			return htonl(NFS4ERR_RETRY_UNCACHED_REP);

		/* The ca_maxresponsesize_cached is 0 with no DRC */
		else if (args->csa_cachethis == 1)
			return htonl(NFS4ERR_REP_TOO_BIG_TO_CACHE);
	}

	/* Wraparound */
	if (args->csa_sequenceid == 1 && (slot->seq_nr + 1) == 0) {
		slot->seq_nr = 1;
		goto out_ok;
	}

	/* Misordered request */
	return htonl(NFS4ERR_SEQ_MISORDERED);
out_ok:
	tbl->highest_used_slotid = args->csa_slotid;
	return htonl(NFS4_OK);
}

/*
 * For each referring call triple, check the session's slot table for
 * a match.  If the slot is in use and the sequence numbers match, the
 * client is still waiting for a response to the original request.
 */
static bool referring_call_exists(struct nfs_client *clp,
				  uint32_t nrclists,
				  struct referring_call_list *rclists)
{
	bool status = 0;
	int i, j;
	struct nfs4_session *session;
	struct nfs4_slot_table *tbl;
	struct referring_call_list *rclist;
	struct referring_call *ref;

	/*
	 * XXX When client trunking is implemented, this becomes
	 * a session lookup from within the loop
	 */
	session = clp->cl_session;
	tbl = &session->fc_slot_table;

	for (i = 0; i < nrclists; i++) {
		rclist = &rclists[i];
		if (memcmp(session->sess_id.data,
			   rclist->rcl_sessionid.data,
			   NFS4_MAX_SESSIONID_LEN) != 0)
			continue;

		for (j = 0; j < rclist->rcl_nrefcalls; j++) {
			ref = &rclist->rcl_refcalls[j];

			dprintk("%s: sessionid %x:%x:%x:%x sequenceid %u "
				"slotid %u\n", __func__,
				((u32 *)&rclist->rcl_sessionid.data)[0],
				((u32 *)&rclist->rcl_sessionid.data)[1],
				((u32 *)&rclist->rcl_sessionid.data)[2],
				((u32 *)&rclist->rcl_sessionid.data)[3],
				ref->rc_sequenceid, ref->rc_slotid);

			spin_lock(&tbl->slot_tbl_lock);
			status = (test_bit(ref->rc_slotid, tbl->used_slots) &&
				  tbl->slots[ref->rc_slotid].seq_nr ==
					ref->rc_sequenceid);
			spin_unlock(&tbl->slot_tbl_lock);
			if (status)
				goto out;
		}
	}

out:
	return status;
}

__be32 nfs4_callback_sequence(struct cb_sequenceargs *args,
			      struct cb_sequenceres *res,
			      struct cb_process_state *cps)
{
	struct nfs4_slot_table *tbl;
	struct nfs4_slot *slot;
	struct nfs_client *clp;
	int i;
	__be32 status = htonl(NFS4ERR_BADSESSION);

	clp = nfs4_find_client_sessionid(cps->net, args->csa_addr,
					 &args->csa_sessionid, cps->minorversion);
	if (clp == NULL)
		goto out;

	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
		goto out;

	tbl = &clp->cl_session->bc_slot_table;
	slot = tbl->slots + args->csa_slotid;

	spin_lock(&tbl->slot_tbl_lock);
	/* state manager is resetting the session */
	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state)) {
		status = htonl(NFS4ERR_DELAY);
		/* Return NFS4ERR_BADSESSION if we're draining the session
		 * in order to reset it.
		 */
		if (test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
			status = htonl(NFS4ERR_BADSESSION);
		goto out_unlock;
	}

	memcpy(&res->csr_sessionid, &args->csa_sessionid,
	       sizeof(res->csr_sessionid));
	res->csr_sequenceid = args->csa_sequenceid;
	res->csr_slotid = args->csa_slotid;
	res->csr_highestslotid = NFS41_BC_MAX_CALLBACKS - 1;
	res->csr_target_highestslotid = NFS41_BC_MAX_CALLBACKS - 1;

	status = validate_seqid(tbl, args);
	if (status)
		goto out_unlock;

	cps->slotid = args->csa_slotid;

	/*
	 * Check for pending referring calls.  If a match is found, a
	 * related callback was received before the response to the original
	 * call.
	 */
	if (referring_call_exists(clp, args->csa_nrclists, args->csa_rclists)) {
		status = htonl(NFS4ERR_DELAY);
		goto out_unlock;
	}

	/*
	 * RFC5661 20.9.3
	 * If CB_SEQUENCE returns an error, then the state of the slot
	 * (sequence ID, cached reply) MUST NOT change.
	 */
	slot->seq_nr++;
out_unlock:
	spin_unlock(&tbl->slot_tbl_lock);

out:
	cps->clp = clp; /* put in nfs4_callback_compound */
	for (i = 0; i < args->csa_nrclists; i++)
		kfree(args->csa_rclists[i].rcl_refcalls);
	kfree(args->csa_rclists);

	if (status == htonl(NFS4ERR_RETRY_UNCACHED_REP)) {
		cps->drc_status = status;
		status = 0;
	} else
		res->csr_status = status;

	trace_nfs4_cb_sequence(args, res, status);
	dprintk("%s: exit with status = %d res->csr_status %d\n", __func__,
		ntohl(status), ntohl(res->csr_status));
	return status;
}

static bool
validate_bitmap_values(unsigned long mask)
{
	return (mask & ~RCA4_TYPE_MASK_ALL) == 0;
}

__be32 nfs4_callback_recallany(struct cb_recallanyargs *args, void *dummy,
			       struct cb_process_state *cps)
{
	__be32 status;
	fmode_t flags = 0;

	status = cpu_to_be32(NFS4ERR_OP_NOT_IN_SESSION);
	if (!cps->clp) /* set in cb_sequence */
		goto out;

	dprintk_rcu("NFS: RECALL_ANY callback request from %s\n",
		rpc_peeraddr2str(cps->clp->cl_rpcclient, RPC_DISPLAY_ADDR));

	status = cpu_to_be32(NFS4ERR_INVAL);
	if (!validate_bitmap_values(args->craa_type_mask))
		goto out;

	status = cpu_to_be32(NFS4_OK);
	if (test_bit(RCA4_TYPE_MASK_RDATA_DLG, (const unsigned long *)
		     &args->craa_type_mask))
		flags = FMODE_READ;
	if (test_bit(RCA4_TYPE_MASK_WDATA_DLG, (const unsigned long *)
		     &args->craa_type_mask))
		flags |= FMODE_WRITE;
	if (test_bit(RCA4_TYPE_MASK_FILE_LAYOUT, (const unsigned long *)
		     &args->craa_type_mask))
		pnfs_recall_all_layouts(cps->clp);
	if (flags)
		nfs_expire_unused_delegation_types(cps->clp, flags);
out:
	dprintk("%s: exit with status = %d\n", __func__, ntohl(status));
	return status;
}

/* Reduce the fore channel's max_slots to the target value */
__be32 nfs4_callback_recallslot(struct cb_recallslotargs *args, void *dummy,
				struct cb_process_state *cps)
{
	struct nfs4_slot_table *fc_tbl;
	__be32 status;

	status = htonl(NFS4ERR_OP_NOT_IN_SESSION);
	if (!cps->clp) /* set in cb_sequence */
		goto out;

	dprintk_rcu("NFS: CB_RECALL_SLOT request from %s target highest slotid %u\n",
		rpc_peeraddr2str(cps->clp->cl_rpcclient, RPC_DISPLAY_ADDR),
		args->crsa_target_highest_slotid);

	fc_tbl = &cps->clp->cl_session->fc_slot_table;

	status = htonl(NFS4_OK);

	nfs41_set_target_slotid(fc_tbl, args->crsa_target_highest_slotid);
	nfs41_notify_server(cps->clp);
out:
	dprintk("%s: exit with status = %d\n", __func__, ntohl(status));
	return status;
}
#endif /* CONFIG_NFS_V4_1 */
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/fs/nfs/callback_proc.c
	3	*
	4	* Copyright (C) 2004 Trond Myklebust
	5	*
	6	* NFSv4 callback procedures
	7	*/
1da177e4 LT	8	#include <linux/nfs4.h>
1da177e4 LT	9	#include <linux/nfs_fs.h>
5a0e3ad6	10	#include <linux/slab.h>
2446ab60	11	#include <linux/rcupdate.h>
4ce79717	12	#include "nfs4_fs.h"
1da177e4 LT	13	#include "callback.h"
1da177e4 LT	14	#include "delegation.h"
24c8dbbb	15	#include "internal.h"
43f1b3da	16	#include "pnfs.h"
73e39aaa	17	#include "nfs4session.h"
ca8acf8d	18	#include "nfs4trace.h"
1da177e4 LT	19
1da177e4 LT	20	#define NFSDBG_FACILITY NFSDBG_CALLBACK
c36fca52 AA	21
	22	__be32 nfs4_callback_getattr(struct cb_getattrargs *args,
	23	struct cb_getattrres *res,
	24	struct cb_process_state *cps)
1da177e4	25	{
1da177e4 LT	26	struct nfs_delegation *delegation;
	27	struct nfs_inode *nfsi;
	28	struct inode *inode;
1d98fe67	29
c36fca52 AA	30	res->status = htonl(NFS4ERR_OP_NOT_IN_SESSION);
	31	if (!cps->clp) /* Always set for v4.0. Set in cb_sequence for v4.1 */
	32	goto out;
	33
1da177e4 LT	34	res->bitmap[0] = res->bitmap[1] = 0;
1da177e4 LT	35	res->status = htonl(NFS4ERR_BADHANDLE);
1d98fe67	36
2446ab60	37	dprintk_rcu("NFS: GETATTR callback request from %s\n",
c36fca52	38	rpc_peeraddr2str(cps->clp->cl_rpcclient, RPC_DISPLAY_ADDR));
1d98fe67	39
c36fca52	40	inode = nfs_delegation_find_inode(cps->clp, &args->fh);
7cd14861	41	if (inode == NULL) {
249b2eef	42	trace_nfs4_cb_getattr(cps->clp, &args->fh, NULL,
7cd14861	43	-ntohl(res->status));
c36fca52	44	goto out;
7cd14861	45	}
1da177e4	46	nfsi = NFS_I(inode);
761fe93c TM	47	rcu_read_lock();
761fe93c TM	48	delegation = rcu_dereference(nfsi->delegation);
1da177e4 LT	49	if (delegation == NULL \|\| (delegation->type & FMODE_WRITE) == 0)
	50	goto out_iput;
	51	res->size = i_size_read(inode);
beb2a5ec	52	res->change_attr = delegation->change_attr;
cb1410c7	53	if (nfsi->nrequests != 0)
beb2a5ec	54	res->change_attr++;
1da177e4 LT	55	res->ctime = inode->i_ctime;
	56	res->mtime = inode->i_mtime;
	57	res->bitmap[0] = (FATTR4_WORD0_CHANGE\|FATTR4_WORD0_SIZE) &
	58	args->bitmap[0];
	59	res->bitmap[1] = (FATTR4_WORD1_TIME_METADATA\|FATTR4_WORD1_TIME_MODIFY) &
	60	args->bitmap[1];
	61	res->status = 0;
	62	out_iput:
761fe93c	63	rcu_read_unlock();
7cd14861	64	trace_nfs4_cb_getattr(cps->clp, &args->fh, inode, -ntohl(res->status));
1da177e4	65	iput(inode);
1da177e4	66	out:
3110ff80	67	dprintk("%s: exit with status = %d\n", __func__, ntohl(res->status));
1da177e4 LT	68	return res->status;
	69	}
	70
c36fca52 AA	71	__be32 nfs4_callback_recall(struct cb_recallargs args, void dummy,
c36fca52 AA	72	struct cb_process_state *cps)
1da177e4	73	{
1da177e4	74	struct inode *inode;
e6f684f6	75	__be32 res;
1da177e4	76
c36fca52 AA	77	res = htonl(NFS4ERR_OP_NOT_IN_SESSION);
c36fca52 AA	78	if (!cps->clp) /* Always set for v4.0. Set in cb_sequence for v4.1 */
1da177e4	79	goto out;
1d98fe67	80
2446ab60	81	dprintk_rcu("NFS: RECALL callback request from %s\n",
c36fca52 AA	82	rpc_peeraddr2str(cps->clp->cl_rpcclient, RPC_DISPLAY_ADDR));
	83
	84	res = htonl(NFS4ERR_BADHANDLE);
	85	inode = nfs_delegation_find_inode(cps->clp, &args->fh);
	86	if (inode == NULL)
	87	goto out;
	88	/* Set up a helper thread to actually return the delegation */
	89	switch (nfs_async_inode_return_delegation(inode, &args->stateid)) {
	90	case 0:
	91	res = 0;
	92	break;
	93	case -ENOENT:
8e663f0e	94	res = htonl(NFS4ERR_BAD_STATEID);
c36fca52 AA	95	break;
	96	default:
	97	res = htonl(NFS4ERR_RESOURCE);
	98	}
ca8acf8d	99	trace_nfs4_recall_delegation(inode, -ntohl(res));
c36fca52	100	iput(inode);
1da177e4	101	out:
3110ff80	102	dprintk("%s: exit with status = %d\n", __func__, ntohl(res));
1da177e4 LT	103	return res;
1da177e4 LT	104	}
d49433e1 BH	105
	106	#if defined(CONFIG_NFS_V4_1)
	107
d36b7cf7 BH	108	/*
	109	* Lookup a layout by filehandle.
	110	*
	111	* Note: gets a refcount on the layout hdr and on its respective inode.
	112	* Caller must put the layout hdr and the inode.
	113	*
	114	* TODO: keep track of all layouts (and delegations) in a hash table
	115	* hashed by filehandle.
	116	*/
27999f25 TM	117	static struct pnfs_layout_hdr * get_layout_by_fh_locked(struct nfs_client *clp,
27999f25 TM	118	struct nfs_fh fh, nfs4_stateid stateid)
43f1b3da	119	{
6382a441	120	struct nfs_server *server;
43f1b3da	121	struct inode *ino;
d36b7cf7	122	struct pnfs_layout_hdr *lo;
43f1b3da	123
6382a441 WAA	124	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
6382a441 WAA	125	list_for_each_entry(lo, &server->layouts, plh_layouts) {
27999f25 TM	126	if (!nfs4_stateid_match_other(&lo->plh_stateid, stateid))
27999f25 TM	127	continue;
d36b7cf7	128	if (nfs_compare_fh(fh, &NFS_I(lo->plh_inode)->fh))
6382a441 WAA	129	continue;
	130	ino = igrab(lo->plh_inode);
	131	if (!ino)
9a7fe9e8	132	break;
9c626381 TM	133	spin_lock(&ino->i_lock);
	134	/* Is this layout in the process of being freed? */
	135	if (NFS_I(ino)->layout != lo) {
	136	spin_unlock(&ino->i_lock);
	137	iput(ino);
9a7fe9e8	138	break;
9c626381	139	}
70c3bd2b	140	pnfs_get_layout_hdr(lo);
9c626381	141	spin_unlock(&ino->i_lock);
d36b7cf7	142	return lo;
6382a441	143	}
43f1b3da	144	}
d36b7cf7 BH	145
	146	return NULL;
	147	}
	148
27999f25 TM	149	static struct pnfs_layout_hdr * get_layout_by_fh(struct nfs_client *clp,
27999f25 TM	150	struct nfs_fh fh, nfs4_stateid stateid)
d36b7cf7 BH	151	{
	152	struct pnfs_layout_hdr *lo;
	153
	154	spin_lock(&clp->cl_lock);
	155	rcu_read_lock();
27999f25	156	lo = get_layout_by_fh_locked(clp, fh, stateid);
6382a441	157	rcu_read_unlock();
43f1b3da	158	spin_unlock(&clp->cl_lock);
6382a441	159
d36b7cf7 BH	160	return lo;
	161	}
	162
41c9127d TM	163	/*
	164	* Enforce RFC5661 section 12.5.5.2.1. (Layout Recall and Return Sequencing)
	165	*/
	166	static bool pnfs_check_stateid_sequence(struct pnfs_layout_hdr *lo,
	167	const nfs4_stateid *new)
	168	{
	169	u32 oldseq, newseq;
	170
	171	oldseq = be32_to_cpu(lo->plh_stateid.seqid);
	172	newseq = be32_to_cpu(new->seqid);
	173
	174	if (newseq > oldseq + 1)
	175	return false;
	176	return true;
	177	}
	178
d36b7cf7 BH	179	static u32 initiate_file_draining(struct nfs_client *clp,
	180	struct cb_layoutrecallargs *args)
	181	{
	182	struct inode *ino;
	183	struct pnfs_layout_hdr *lo;
	184	u32 rv = NFS4ERR_NOMATCHING_LAYOUT;
	185	LIST_HEAD(free_me_list);
	186
27999f25	187	lo = get_layout_by_fh(clp, &args->cbl_fh, &args->cbl_stateid);
d36b7cf7	188	if (!lo)
27999f25	189	goto out;
43f1b3da	190
d36b7cf7	191	ino = lo->plh_inode;
7c5d1875 CH	192
7c5d1875 CH	193	spin_lock(&ino->i_lock);
41c9127d TM	194	if (!pnfs_check_stateid_sequence(lo, &args->cbl_stateid)) {
	195	rv = NFS4ERR_DELAY;
	196	goto unlock;
	197	}
7c5d1875 CH	198	pnfs_set_layout_stateid(lo, &args->cbl_stateid, true);
	199	spin_unlock(&ino->i_lock);
	200
	201	pnfs_layoutcommit_inode(ino, false);
	202
43f1b3da	203	spin_lock(&ino->i_lock);
fc7ff367 TM	204	/*
	205	* Enforce RFC5661 Section 12.5.5.2.1.5 (Bulk Recall and Return)
	206	*/
	207	if (test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags)) {
	208	rv = NFS4ERR_DELAY;
	209	goto unlock;
	210	}
	211
	212	if (pnfs_mark_matching_lsegs_invalid(lo, &free_me_list,
c88953d8	213	&args->cbl_range)) {
fc7ff367 TM	214	pnfs_mark_matching_lsegs_return(lo,
	215	&free_me_list,
	216	&args->cbl_range);
43f1b3da	217	rv = NFS4ERR_DELAY;
c88953d8 CH	218	goto unlock;
	219	}
	220
	221	if (NFS_SERVER(ino)->pnfs_curr_ld->return_range) {
	222	NFS_SERVER(ino)->pnfs_curr_ld->return_range(lo,
	223	&args->cbl_range);
	224	}
0654cc72	225	pnfs_mark_layout_returned_if_empty(lo);
c88953d8	226	unlock:
43f1b3da FI	227	spin_unlock(&ino->i_lock);
43f1b3da FI	228	pnfs_free_lseg_list(&free_me_list);
70c3bd2b	229	pnfs_put_layout_hdr(lo);
249b2eef	230	trace_nfs4_cb_layoutrecall_inode(clp, &args->cbl_fh, ino, -rv);
43f1b3da	231	iput(ino);
27999f25	232	out:
43f1b3da FI	233	return rv;
	234	}
	235
	236	static u32 initiate_bulk_draining(struct nfs_client *clp,
	237	struct cb_layoutrecallargs *args)
	238	{
fd9a8d71	239	int stat;
6382a441	240
fd9a8d71 TM	241	if (args->cbl_recall_type == RETURN_FSID)
	242	stat = pnfs_destroy_layouts_byfsid(clp, &args->cbl_fsid, true);
	243	else
	244	stat = pnfs_destroy_layouts_byclid(clp, true);
	245	if (stat != 0)
	246	return NFS4ERR_DELAY;
	247	return NFS4ERR_NOMATCHING_LAYOUT;
43f1b3da FI	248	}
	249
	250	static u32 do_callback_layoutrecall(struct nfs_client *clp,
	251	struct cb_layoutrecallargs *args)
	252	{
0cb3284b	253	u32 res;
43f1b3da FI	254
43f1b3da FI	255	dprintk("%s enter, type=%i\n", __func__, args->cbl_recall_type);
43f1b3da FI	256	if (args->cbl_recall_type == RETURN_FILE)
	257	res = initiate_file_draining(clp, args);
	258	else
	259	res = initiate_bulk_draining(clp, args);
43f1b3da FI	260	dprintk("%s returning %i\n", __func__, res);
	261	return res;
	262
	263	}
	264
f2a62561 FI	265	__be32 nfs4_callback_layoutrecall(struct cb_layoutrecallargs *args,
	266	void dummy, struct cb_process_state cps)
	267	{
43f1b3da FI	268	u32 res;
	269
	270	dprintk("%s: -->\n", __func__);
	271
	272	if (cps->clp)
	273	res = do_callback_layoutrecall(cps->clp, args);
	274	else
	275	res = NFS4ERR_OP_NOT_IN_SESSION;
	276
	277	dprintk("%s: exit with status = %d\n", __func__, res);
	278	return cpu_to_be32(res);
f2a62561 FI	279	}
f2a62561 FI	280
36840370 AB	281	static void pnfs_recall_all_layouts(struct nfs_client *clp)
	282	{
	283	struct cb_layoutrecallargs args;
	284
	285	/* Pretend we got a CB_LAYOUTRECALL(ALL) */
	286	memset(&args, 0, sizeof(args));
	287	args.cbl_recall_type = RETURN_ALL;
	288	/* FIXME we ignore errors, what should we do? */
	289	do_callback_layoutrecall(clp, &args);
	290	}
	291
1be5683b ME	292	__be32 nfs4_callback_devicenotify(struct cb_devicenotifyargs *args,
	293	void dummy, struct cb_process_state cps)
	294	{
	295	int i;
	296	__be32 res = 0;
	297	struct nfs_client *clp = cps->clp;
	298	struct nfs_server *server = NULL;
	299
	300	dprintk("%s: -->\n", __func__);
	301
	302	if (!clp) {
	303	res = cpu_to_be32(NFS4ERR_OP_NOT_IN_SESSION);
	304	goto out;
	305	}
	306
	307	for (i = 0; i < args->ndevs; i++) {
	308	struct cb_devicenotifyitem *dev = &args->devs[i];
	309
	310	if (!server \|\|
	311	server->pnfs_curr_ld->id != dev->cbd_layout_type) {
	312	rcu_read_lock();
	313	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
	314	if (server->pnfs_curr_ld &&
	315	server->pnfs_curr_ld->id == dev->cbd_layout_type) {
	316	rcu_read_unlock();
	317	goto found;
	318	}
	319	rcu_read_unlock();
	320	dprintk("%s: layout type %u not found\n",
	321	__func__, dev->cbd_layout_type);
	322	continue;
	323	}
	324
	325	found:
35c8bb54	326	nfs4_delete_deviceid(server->pnfs_curr_ld, clp, &dev->cbd_dev_id);
1be5683b ME	327	}
	328
	329	out:
	330	kfree(args->devs);
	331	dprintk("%s: exit with status = %u\n",
	332	__func__, be32_to_cpu(res));
	333	return res;
	334	}
	335
68f3f901 RL	336	/*
	337	* Validate the sequenceID sent by the server.
	338	* Return success if the sequenceID is one more than what we last saw on
	339	* this slot, accounting for wraparound. Increments the slot's sequence.
	340	*
4911096f AA	341	* We don't yet implement a duplicate request cache, instead we set the
4911096f AA	342	* back channel ca_maxresponsesize_cached to zero. This is OK for now
68f3f901 RL	343	* since we only currently implement idempotent callbacks anyway.
	344	*
	345	* We have a single slot backchannel at this time, so we don't bother
	346	* checking the used_slots bit array on the table. The lower layer guarantees
	347	* a single outstanding callback request at a time.
	348	*/
9733f0d9	349	static __be32
b2f28bd7	350	validate_seqid(struct nfs4_slot_table tbl, struct cb_sequenceargs args)
68f3f901 RL	351	{
	352	struct nfs4_slot *slot;
	353
e8d92382	354	dprintk("%s enter. slotid %u seqid %u\n",
b2f28bd7	355	__func__, args->csa_slotid, args->csa_sequenceid);
68f3f901	356
61f2e510	357	if (args->csa_slotid >= NFS41_BC_MAX_CALLBACKS)
68f3f901 RL	358	return htonl(NFS4ERR_BADSLOT);
68f3f901 RL	359
b2f28bd7	360	slot = tbl->slots + args->csa_slotid;
e8d92382	361	dprintk("%s slot table seqid: %u\n", __func__, slot->seq_nr);
68f3f901 RL	362
68f3f901 RL	363	/* Normal */
e937ee71	364	if (likely(args->csa_sequenceid == slot->seq_nr + 1))
55a67399	365	goto out_ok;
68f3f901 RL	366
68f3f901 RL	367	/* Replay */
b2f28bd7	368	if (args->csa_sequenceid == slot->seq_nr) {
e8d92382	369	dprintk("%s seqid %u is a replay\n",
b2f28bd7	370	__func__, args->csa_sequenceid);
4911096f AA	371	/* Signal process_op to set this error on next op */
	372	if (args->csa_cachethis == 0)
	373	return htonl(NFS4ERR_RETRY_UNCACHED_REP);
	374
	375	/* The ca_maxresponsesize_cached is 0 with no DRC */
	376	else if (args->csa_cachethis == 1)
	377	return htonl(NFS4ERR_REP_TOO_BIG_TO_CACHE);
68f3f901 RL	378	}
	379
	380	/* Wraparound */
b2f28bd7	381	if (args->csa_sequenceid == 1 && (slot->seq_nr + 1) == 0) {
68f3f901	382	slot->seq_nr = 1;
55a67399	383	goto out_ok;
68f3f901 RL	384	}
	385
	386	/* Misordered request */
	387	return htonl(NFS4ERR_SEQ_MISORDERED);
55a67399 TM	388	out_ok:
	389	tbl->highest_used_slotid = args->csa_slotid;
	390	return htonl(NFS4_OK);
68f3f901 RL	391	}
68f3f901 RL	392
a7989c3e MS	393	/*
	394	* For each referring call triple, check the session's slot table for
	395	* a match. If the slot is in use and the sequence numbers match, the
	396	* client is still waiting for a response to the original request.
	397	*/
	398	static bool referring_call_exists(struct nfs_client *clp,
	399	uint32_t nrclists,
	400	struct referring_call_list *rclists)
	401	{
	402	bool status = 0;
	403	int i, j;
	404	struct nfs4_session *session;
	405	struct nfs4_slot_table *tbl;
	406	struct referring_call_list *rclist;
	407	struct referring_call *ref;
	408
	409	/*
	410	* XXX When client trunking is implemented, this becomes
	411	* a session lookup from within the loop
	412	*/
	413	session = clp->cl_session;
	414	tbl = &session->fc_slot_table;
	415
	416	for (i = 0; i < nrclists; i++) {
	417	rclist = &rclists[i];
	418	if (memcmp(session->sess_id.data,
	419	rclist->rcl_sessionid.data,
	420	NFS4_MAX_SESSIONID_LEN) != 0)
	421	continue;
	422
	423	for (j = 0; j < rclist->rcl_nrefcalls; j++) {
	424	ref = &rclist->rcl_refcalls[j];
	425
	426	dprintk("%s: sessionid %x:%x:%x:%x sequenceid %u "
	427	"slotid %u\n", __func__,
	428	((u32 *)&rclist->rcl_sessionid.data)[0],
	429	((u32 *)&rclist->rcl_sessionid.data)[1],
	430	((u32 *)&rclist->rcl_sessionid.data)[2],
	431	((u32 *)&rclist->rcl_sessionid.data)[3],
	432	ref->rc_sequenceid, ref->rc_slotid);
	433
	434	spin_lock(&tbl->slot_tbl_lock);
	435	status = (test_bit(ref->rc_slotid, tbl->used_slots) &&
	436	tbl->slots[ref->rc_slotid].seq_nr ==
	437	ref->rc_sequenceid);
	438	spin_unlock(&tbl->slot_tbl_lock);
	439	if (status)
	440	goto out;
	441	}
	442	}
	443
	444	out:
	445	return status;
	446	}
	447
9733f0d9	448	__be32 nfs4_callback_sequence(struct cb_sequenceargs *args,
c36fca52 AA	449	struct cb_sequenceres *res,
c36fca52 AA	450	struct cb_process_state *cps)
d49433e1	451	{
55a67399	452	struct nfs4_slot_table *tbl;
e937ee71	453	struct nfs4_slot *slot;
68f3f901	454	struct nfs_client *clp;
9733f0d9	455	int i;
778be232	456	__be32 status = htonl(NFS4ERR_BADSESSION);
d49433e1	457
459de2ed BS	458	clp = nfs4_find_client_sessionid(cps->net, args->csa_addr,
459de2ed BS	459	&args->csa_sessionid, cps->minorversion);
68f3f901 RL	460	if (clp == NULL)
	461	goto out;
	462
b1c0df5f TM	463	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
b1c0df5f TM	464	goto out;
e937ee71	465
55a67399	466	tbl = &clp->cl_session->bc_slot_table;
e937ee71	467	slot = tbl->slots + args->csa_slotid;
55a67399 TM	468
55a67399 TM	469	spin_lock(&tbl->slot_tbl_lock);
42acd021	470	/* state manager is resetting the session */
774d5f14	471	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state)) {
55a67399	472	status = htonl(NFS4ERR_DELAY);
910ac68a TM	473	/* Return NFS4ERR_BADSESSION if we're draining the session
	474	* in order to reset it.
	475	*/
	476	if (test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
	477	status = htonl(NFS4ERR_BADSESSION);
4e54ab8d	478	goto out_unlock;
42acd021 AA	479	}
42acd021 AA	480
0579c8d2 KM	481	memcpy(&res->csr_sessionid, &args->csa_sessionid,
	482	sizeof(res->csr_sessionid));
	483	res->csr_sequenceid = args->csa_sequenceid;
	484	res->csr_slotid = args->csa_slotid;
	485	res->csr_highestslotid = NFS41_BC_MAX_CALLBACKS - 1;
	486	res->csr_target_highestslotid = NFS41_BC_MAX_CALLBACKS - 1;
	487
e937ee71	488	status = validate_seqid(tbl, args);
68f3f901	489	if (status)
4e54ab8d	490	goto out_unlock;
68f3f901	491
55a67399 TM	492	cps->slotid = args->csa_slotid;
55a67399 TM	493
72ce2b3c MS	494	/*
	495	* Check for pending referring calls. If a match is found, a
	496	* related callback was received before the response to the original
	497	* call.
	498	*/
	499	if (referring_call_exists(clp, args->csa_nrclists, args->csa_rclists)) {
	500	status = htonl(NFS4ERR_DELAY);
4e54ab8d	501	goto out_unlock;
72ce2b3c MS	502	}
72ce2b3c MS	503
e937ee71 KM	504	/*
	505	* RFC5661 20.9.3
	506	* If CB_SEQUENCE returns an error, then the state of the slot
	507	* (sequence ID, cached reply) MUST NOT change.
	508	*/
	509	slot->seq_nr++;
4e54ab8d TM	510	out_unlock:
4e54ab8d TM	511	spin_unlock(&tbl->slot_tbl_lock);
e937ee71	512
68f3f901	513	out:
2c4cdf8f	514	cps->clp = clp; /* put in nfs4_callback_compound */
72ce2b3c MS	515	for (i = 0; i < args->csa_nrclists; i++)
	516	kfree(args->csa_rclists[i].rcl_refcalls);
	517	kfree(args->csa_rclists);
	518
c36fca52 AA	519	if (status == htonl(NFS4ERR_RETRY_UNCACHED_REP)) {
	520	cps->drc_status = status;
	521	status = 0;
	522	} else
4911096f	523	res->csr_status = status;
c36fca52	524
2f92ae34	525	trace_nfs4_cb_sequence(args, res, status);
4911096f AA	526	dprintk("%s: exit with status = %d res->csr_status %d\n", __func__,
	527	ntohl(status), ntohl(res->csr_status));
	528	return status;
d49433e1 BH	529	}
d49433e1 BH	530
36840370 AB	531	static bool
	532	validate_bitmap_values(unsigned long mask)
	533	{
	534	return (mask & ~RCA4_TYPE_MASK_ALL) == 0;
	535	}
	536
c36fca52 AA	537	__be32 nfs4_callback_recallany(struct cb_recallanyargs args, void dummy,
c36fca52 AA	538	struct cb_process_state *cps)
31f09607	539	{
9733f0d9	540	__be32 status;
31f09607 AB	541	fmode_t flags = 0;
31f09607 AB	542
36840370	543	status = cpu_to_be32(NFS4ERR_OP_NOT_IN_SESSION);
c36fca52	544	if (!cps->clp) /* set in cb_sequence */
31f09607 AB	545	goto out;
31f09607 AB	546
2446ab60	547	dprintk_rcu("NFS: RECALL_ANY callback request from %s\n",
c36fca52	548	rpc_peeraddr2str(cps->clp->cl_rpcclient, RPC_DISPLAY_ADDR));
31f09607	549
36840370 AB	550	status = cpu_to_be32(NFS4ERR_INVAL);
	551	if (!validate_bitmap_values(args->craa_type_mask))
	552	goto out;
	553
	554	status = cpu_to_be32(NFS4_OK);
31f09607 AB	555	if (test_bit(RCA4_TYPE_MASK_RDATA_DLG, (const unsigned long *)
	556	&args->craa_type_mask))
	557	flags = FMODE_READ;
	558	if (test_bit(RCA4_TYPE_MASK_WDATA_DLG, (const unsigned long *)
	559	&args->craa_type_mask))
	560	flags \|= FMODE_WRITE;
36840370 AB	561	if (test_bit(RCA4_TYPE_MASK_FILE_LAYOUT, (const unsigned long *)
	562	&args->craa_type_mask))
	563	pnfs_recall_all_layouts(cps->clp);
31f09607	564	if (flags)
826e0013	565	nfs_expire_unused_delegation_types(cps->clp, flags);
31f09607 AB	566	out:
	567	dprintk("%s: exit with status = %d\n", __func__, ntohl(status));
	568	return status;
	569	}
b9efa1b2 AA	570
b9efa1b2 AA	571	/* Reduce the fore channel's max_slots to the target value */
c36fca52 AA	572	__be32 nfs4_callback_recallslot(struct cb_recallslotargs args, void dummy,
c36fca52 AA	573	struct cb_process_state *cps)
b9efa1b2	574	{
b9efa1b2	575	struct nfs4_slot_table *fc_tbl;
9733f0d9	576	__be32 status;
b9efa1b2 AA	577
b9efa1b2 AA	578	status = htonl(NFS4ERR_OP_NOT_IN_SESSION);
c36fca52	579	if (!cps->clp) /* set in cb_sequence */
b9efa1b2 AA	580	goto out;
b9efa1b2 AA	581
e8d92382	582	dprintk_rcu("NFS: CB_RECALL_SLOT request from %s target highest slotid %u\n",
c36fca52	583	rpc_peeraddr2str(cps->clp->cl_rpcclient, RPC_DISPLAY_ADDR),
d5fb4ce3	584	args->crsa_target_highest_slotid);
b9efa1b2	585
c36fca52	586	fc_tbl = &cps->clp->cl_session->fc_slot_table;
b9efa1b2	587
bae0ac0e	588	status = htonl(NFS4_OK);
b9efa1b2	589
d5fb4ce3	590	nfs41_set_target_slotid(fc_tbl, args->crsa_target_highest_slotid);
3f10a6af	591	nfs41_notify_server(cps->clp);
b9efa1b2 AA	592	out:
	593	dprintk("%s: exit with status = %d\n", __func__, ntohl(status));
	594	return status;
	595	}
d49433e1	596	#endif /* CONFIG_NFS_V4_1 */