[deliverable/linux.git] / drivers / staging / lustre / lustre / mdc / mdc_lib.c

/*
 * GPL HEADER START
 *
 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 only,
 * as published by the Free Software Foundation.
 *
 * 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 version 2 for more details (a copy is included
 * in the LICENSE file that accompanied this code).
 *
 * You should have received a copy of the GNU General Public License
 * version 2 along with this program; If not, see
 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
 *
 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
 * CA 95054 USA or visit www.sun.com if you need additional information or
 * have any questions.
 *
 * GPL HEADER END
 */
/*
 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
 * Use is subject to license terms.
 *
 * Copyright (c) 2011, 2012, Intel Corporation.
 */
/*
 * This file is part of Lustre, http://www.lustre.org/
 * Lustre is a trademark of Sun Microsystems, Inc.
 */

#define DEBUG_SUBSYSTEM S_MDC
#include <lustre_net.h>
#include <lustre/lustre_idl.h>
#include "mdc_internal.h"


static void __mdc_pack_body(struct mdt_body *b, __u32 suppgid)
{
	LASSERT (b != NULL);

	b->suppgid = suppgid;
	b->uid = from_kuid(&init_user_ns, current_uid());
	b->gid = from_kgid(&init_user_ns, current_gid());
	b->fsuid = from_kuid(&init_user_ns, current_fsuid());
	b->fsgid = from_kgid(&init_user_ns, current_fsgid());
	b->capability = cfs_curproc_cap_pack();
}

void mdc_pack_capa(struct ptlrpc_request *req, const struct req_msg_field *field,
		   struct obd_capa *oc)
{
	struct req_capsule *pill = &req->rq_pill;
	struct lustre_capa *c;

	if (oc == NULL) {
		LASSERT(req_capsule_get_size(pill, field, RCL_CLIENT) == 0);
		return;
	}

	c = req_capsule_client_get(pill, field);
	LASSERT(c != NULL);
	capa_cpy(c, oc);
	DEBUG_CAPA(D_SEC, c, "pack");
}

void mdc_is_subdir_pack(struct ptlrpc_request *req, const struct lu_fid *pfid,
			const struct lu_fid *cfid, int flags)
{
	struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
						    &RMF_MDT_BODY);

	if (pfid) {
		b->fid1 = *pfid;
		b->valid = OBD_MD_FLID;
	}
	if (cfid)
		b->fid2 = *cfid;
	b->flags = flags;
}

void mdc_swap_layouts_pack(struct ptlrpc_request *req,
			   struct md_op_data *op_data)
{
	struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
						    &RMF_MDT_BODY);

	__mdc_pack_body(b, op_data->op_suppgids[0]);
	b->fid1 = op_data->op_fid1;
	b->fid2 = op_data->op_fid2;
	b->valid |= OBD_MD_FLID;

	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
	mdc_pack_capa(req, &RMF_CAPA2, op_data->op_capa2);
}

void mdc_pack_body(struct ptlrpc_request *req,
		   const struct lu_fid *fid, struct obd_capa *oc,
		   __u64 valid, int ea_size, __u32 suppgid, int flags)
{
	struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
						    &RMF_MDT_BODY);
	LASSERT(b != NULL);
	b->valid = valid;
	b->eadatasize = ea_size;
	b->flags = flags;
	__mdc_pack_body(b, suppgid);
	if (fid) {
		b->fid1 = *fid;
		b->valid |= OBD_MD_FLID;
		mdc_pack_capa(req, &RMF_CAPA1, oc);
	}
}

void mdc_readdir_pack(struct ptlrpc_request *req, __u64 pgoff,
		      __u32 size, const struct lu_fid *fid, struct obd_capa *oc)
{
	struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
						    &RMF_MDT_BODY);
	b->fid1 = *fid;
	b->valid |= OBD_MD_FLID;
	b->size = pgoff;		       /* !! */
	b->nlink = size;			/* !! */
	__mdc_pack_body(b, -1);
	b->mode = LUDA_FID | LUDA_TYPE;

	mdc_pack_capa(req, &RMF_CAPA1, oc);
}

/* packing of MDS records */
void mdc_create_pack(struct ptlrpc_request *req, struct md_op_data *op_data,
		     const void *data, int datalen, __u32 mode,
		     __u32 uid, __u32 gid, cfs_cap_t cap_effective, __u64 rdev)
{
	struct mdt_rec_create	*rec;
	char			*tmp;
	__u64			 flags;

	CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_create));
	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);


	rec->cr_opcode   = REINT_CREATE;
	rec->cr_fsuid    = uid;
	rec->cr_fsgid    = gid;
	rec->cr_cap      = cap_effective;
	rec->cr_fid1     = op_data->op_fid1;
	rec->cr_fid2     = op_data->op_fid2;
	rec->cr_mode     = mode;
	rec->cr_rdev     = rdev;
	rec->cr_time     = op_data->op_mod_time;
	rec->cr_suppgid1 = op_data->op_suppgids[0];
	rec->cr_suppgid2 = op_data->op_suppgids[1];
	flags = op_data->op_flags & MF_SOM_LOCAL_FLAGS;
	if (op_data->op_bias & MDS_CREATE_VOLATILE)
		flags |= MDS_OPEN_VOLATILE;
	set_mrc_cr_flags(rec, flags);
	rec->cr_bias     = op_data->op_bias;
	rec->cr_umask    = current_umask();

	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);

	tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
	LOGL0(op_data->op_name, op_data->op_namelen, tmp);

	if (data) {
		tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
		memcpy(tmp, data, datalen);
	}
}

static __u64 mds_pack_open_flags(__u32 flags, __u32 mode)
{
	__u64 cr_flags = (flags & (FMODE_READ | FMODE_WRITE |
				   MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS |
				   MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK |
				   MDS_OPEN_BY_FID));
	if (flags & O_CREAT)
		cr_flags |= MDS_OPEN_CREAT;
	if (flags & O_EXCL)
		cr_flags |= MDS_OPEN_EXCL;
	if (flags & O_TRUNC)
		cr_flags |= MDS_OPEN_TRUNC;
	if (flags & O_APPEND)
		cr_flags |= MDS_OPEN_APPEND;
	if (flags & O_SYNC)
		cr_flags |= MDS_OPEN_SYNC;
	if (flags & O_DIRECTORY)
		cr_flags |= MDS_OPEN_DIRECTORY;
#ifdef FMODE_EXEC
	if (flags & FMODE_EXEC)
		cr_flags |= MDS_FMODE_EXEC;
#endif
	if (flags & O_LOV_DELAY_CREATE)
		cr_flags |= MDS_OPEN_DELAY_CREATE;

	if (flags & O_NONBLOCK)
		cr_flags |= MDS_OPEN_NORESTORE;

	return cr_flags;
}

/* packing of MDS records */
void mdc_open_pack(struct ptlrpc_request *req, struct md_op_data *op_data,
		   __u32 mode, __u64 rdev, __u32 flags, const void *lmm,
		   int lmmlen)
{
	struct mdt_rec_create *rec;
	char *tmp;
	__u64 cr_flags;

	CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_create));
	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);

	/* XXX do something about time, uid, gid */
	rec->cr_opcode   = REINT_OPEN;
	rec->cr_fsuid    = from_kuid(&init_user_ns, current_fsuid());
	rec->cr_fsgid    = from_kgid(&init_user_ns, current_fsgid());
	rec->cr_cap      = cfs_curproc_cap_pack();
	if (op_data != NULL) {
		rec->cr_fid1 = op_data->op_fid1;
		rec->cr_fid2 = op_data->op_fid2;
	}
	rec->cr_mode     = mode;
	cr_flags = mds_pack_open_flags(flags, mode);
	rec->cr_rdev     = rdev;
	rec->cr_time     = op_data->op_mod_time;
	rec->cr_suppgid1 = op_data->op_suppgids[0];
	rec->cr_suppgid2 = op_data->op_suppgids[1];
	rec->cr_bias     = op_data->op_bias;
	rec->cr_umask    = current_umask();

	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
	/* the next buffer is child capa, which is used for replay,
	 * will be packed from the data in reply message. */

	if (op_data->op_name) {
		tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
		LOGL0(op_data->op_name, op_data->op_namelen, tmp);
		if (op_data->op_bias & MDS_CREATE_VOLATILE)
			cr_flags |= MDS_OPEN_VOLATILE;
	}

	if (lmm) {
		cr_flags |= MDS_OPEN_HAS_EA;
		tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
		memcpy(tmp, lmm, lmmlen);
	}
	set_mrc_cr_flags(rec, cr_flags);
}

static inline __u64 attr_pack(unsigned int ia_valid) {
	__u64 sa_valid = 0;

	if (ia_valid & ATTR_MODE)
		sa_valid |= MDS_ATTR_MODE;
	if (ia_valid & ATTR_UID)
		sa_valid |= MDS_ATTR_UID;
	if (ia_valid & ATTR_GID)
		sa_valid |= MDS_ATTR_GID;
	if (ia_valid & ATTR_SIZE)
		sa_valid |= MDS_ATTR_SIZE;
	if (ia_valid & ATTR_ATIME)
		sa_valid |= MDS_ATTR_ATIME;
	if (ia_valid & ATTR_MTIME)
		sa_valid |= MDS_ATTR_MTIME;
	if (ia_valid & ATTR_CTIME)
		sa_valid |= MDS_ATTR_CTIME;
	if (ia_valid & ATTR_ATIME_SET)
		sa_valid |= MDS_ATTR_ATIME_SET;
	if (ia_valid & ATTR_MTIME_SET)
		sa_valid |= MDS_ATTR_MTIME_SET;
	if (ia_valid & ATTR_FORCE)
		sa_valid |= MDS_ATTR_FORCE;
	if (ia_valid & ATTR_ATTR_FLAG)
		sa_valid |= MDS_ATTR_ATTR_FLAG;
	if (ia_valid & ATTR_KILL_SUID)
		sa_valid |=  MDS_ATTR_KILL_SUID;
	if (ia_valid & ATTR_KILL_SGID)
		sa_valid |= MDS_ATTR_KILL_SGID;
	if (ia_valid & ATTR_CTIME_SET)
		sa_valid |= MDS_ATTR_CTIME_SET;
	if (ia_valid & ATTR_FROM_OPEN)
		sa_valid |= MDS_ATTR_FROM_OPEN;
	if (ia_valid & ATTR_BLOCKS)
		sa_valid |= MDS_ATTR_BLOCKS;
	if (ia_valid & MDS_OPEN_OWNEROVERRIDE)
		/* NFSD hack (see bug 5781) */
		sa_valid |= MDS_OPEN_OWNEROVERRIDE;
	return sa_valid;
}

static void mdc_setattr_pack_rec(struct mdt_rec_setattr *rec,
				 struct md_op_data *op_data)
{
	rec->sa_opcode  = REINT_SETATTR;
	rec->sa_fsuid   = from_kuid(&init_user_ns, current_fsuid());
	rec->sa_fsgid   = from_kgid(&init_user_ns, current_fsgid());
	rec->sa_cap     = cfs_curproc_cap_pack();
	rec->sa_suppgid = -1;

	rec->sa_fid    = op_data->op_fid1;
	rec->sa_valid  = attr_pack(op_data->op_attr.ia_valid);
	rec->sa_mode   = op_data->op_attr.ia_mode;
	rec->sa_uid    = from_kuid(&init_user_ns, op_data->op_attr.ia_uid);
	rec->sa_gid    = from_kgid(&init_user_ns, op_data->op_attr.ia_gid);
	rec->sa_size   = op_data->op_attr.ia_size;
	rec->sa_blocks = op_data->op_attr_blocks;
	rec->sa_atime  = LTIME_S(op_data->op_attr.ia_atime);
	rec->sa_mtime  = LTIME_S(op_data->op_attr.ia_mtime);
	rec->sa_ctime  = LTIME_S(op_data->op_attr.ia_ctime);
	rec->sa_attr_flags = ((struct ll_iattr *)&op_data->op_attr)->ia_attr_flags;
	if ((op_data->op_attr.ia_valid & ATTR_GID) &&
	    in_group_p(op_data->op_attr.ia_gid))
		rec->sa_suppgid =
			from_kgid(&init_user_ns, op_data->op_attr.ia_gid);
	else
		rec->sa_suppgid = op_data->op_suppgids[0];

	rec->sa_bias = op_data->op_bias;
}

static void mdc_ioepoch_pack(struct mdt_ioepoch *epoch,
			     struct md_op_data *op_data)
{
	memcpy(&epoch->handle, &op_data->op_handle, sizeof(epoch->handle));
	epoch->ioepoch = op_data->op_ioepoch;
	epoch->flags = op_data->op_flags & MF_SOM_LOCAL_FLAGS;
}

void mdc_setattr_pack(struct ptlrpc_request *req, struct md_op_data *op_data,
		      void *ea, int ealen, void *ea2, int ea2len)
{
	struct mdt_rec_setattr *rec;
	struct mdt_ioepoch *epoch;
	struct lov_user_md *lum = NULL;

	CLASSERT(sizeof(struct mdt_rec_reint) ==sizeof(struct mdt_rec_setattr));
	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
	mdc_setattr_pack_rec(rec, op_data);

	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);

	if (op_data->op_flags & (MF_SOM_CHANGE | MF_EPOCH_OPEN)) {
		epoch = req_capsule_client_get(&req->rq_pill, &RMF_MDT_EPOCH);
		mdc_ioepoch_pack(epoch, op_data);
	}

	if (ealen == 0)
		return;

	lum = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
	if (ea == NULL) { /* Remove LOV EA */
		lum->lmm_magic = LOV_USER_MAGIC_V1;
		lum->lmm_stripe_size = 0;
		lum->lmm_stripe_count = 0;
		lum->lmm_stripe_offset = (typeof(lum->lmm_stripe_offset))(-1);
	} else {
		memcpy(lum, ea, ealen);
	}

	if (ea2len == 0)
		return;

	memcpy(req_capsule_client_get(&req->rq_pill, &RMF_LOGCOOKIES), ea2,
	       ea2len);
}

void mdc_unlink_pack(struct ptlrpc_request *req, struct md_op_data *op_data)
{
	struct mdt_rec_unlink *rec;
	char *tmp;

	CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_unlink));
	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
	LASSERT(rec != NULL);

	rec->ul_opcode  = op_data->op_cli_flags & CLI_RM_ENTRY ?
					REINT_RMENTRY : REINT_UNLINK;
	rec->ul_fsuid   = op_data->op_fsuid;
	rec->ul_fsgid   = op_data->op_fsgid;
	rec->ul_cap     = op_data->op_cap;
	rec->ul_mode    = op_data->op_mode;
	rec->ul_suppgid1= op_data->op_suppgids[0];
	rec->ul_suppgid2= -1;
	rec->ul_fid1    = op_data->op_fid1;
	rec->ul_fid2    = op_data->op_fid2;
	rec->ul_time    = op_data->op_mod_time;
	rec->ul_bias    = op_data->op_bias;

	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);

	tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
	LASSERT(tmp != NULL);
	LOGL0(op_data->op_name, op_data->op_namelen, tmp);
}

void mdc_link_pack(struct ptlrpc_request *req, struct md_op_data *op_data)
{
	struct mdt_rec_link *rec;
	char *tmp;

	CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_link));
	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
	LASSERT (rec != NULL);

	rec->lk_opcode   = REINT_LINK;
	rec->lk_fsuid    = op_data->op_fsuid;//current->fsuid;
	rec->lk_fsgid    = op_data->op_fsgid;//current->fsgid;
	rec->lk_cap      = op_data->op_cap;//current->cap_effective;
	rec->lk_suppgid1 = op_data->op_suppgids[0];
	rec->lk_suppgid2 = op_data->op_suppgids[1];
	rec->lk_fid1     = op_data->op_fid1;
	rec->lk_fid2     = op_data->op_fid2;
	rec->lk_time     = op_data->op_mod_time;
	rec->lk_bias     = op_data->op_bias;

	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
	mdc_pack_capa(req, &RMF_CAPA2, op_data->op_capa2);

	tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
	LOGL0(op_data->op_name, op_data->op_namelen, tmp);
}

void mdc_rename_pack(struct ptlrpc_request *req, struct md_op_data *op_data,
		     const char *old, int oldlen, const char *new, int newlen)
{
	struct mdt_rec_rename *rec;
	char *tmp;

	CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_rename));
	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);

	/* XXX do something about time, uid, gid */
	rec->rn_opcode   = REINT_RENAME;
	rec->rn_fsuid    = op_data->op_fsuid;
	rec->rn_fsgid    = op_data->op_fsgid;
	rec->rn_cap      = op_data->op_cap;
	rec->rn_suppgid1 = op_data->op_suppgids[0];
	rec->rn_suppgid2 = op_data->op_suppgids[1];
	rec->rn_fid1     = op_data->op_fid1;
	rec->rn_fid2     = op_data->op_fid2;
	rec->rn_time     = op_data->op_mod_time;
	rec->rn_mode     = op_data->op_mode;
	rec->rn_bias     = op_data->op_bias;

	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
	mdc_pack_capa(req, &RMF_CAPA2, op_data->op_capa2);

	tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
	LOGL0(old, oldlen, tmp);

	if (new) {
		tmp = req_capsule_client_get(&req->rq_pill, &RMF_SYMTGT);
		LOGL0(new, newlen, tmp);
	}
}

void mdc_getattr_pack(struct ptlrpc_request *req, __u64 valid, int flags,
		      struct md_op_data *op_data, int ea_size)
{
	struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
						    &RMF_MDT_BODY);

	b->valid = valid;
	if (op_data->op_bias & MDS_CHECK_SPLIT)
		b->valid |= OBD_MD_FLCKSPLIT;
	if (op_data->op_bias & MDS_CROSS_REF)
		b->valid |= OBD_MD_FLCROSSREF;
	b->eadatasize = ea_size;
	b->flags = flags;
	__mdc_pack_body(b, op_data->op_suppgids[0]);

	b->fid1 = op_data->op_fid1;
	b->fid2 = op_data->op_fid2;
	b->valid |= OBD_MD_FLID;

	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);

	if (op_data->op_name) {
		char *tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
		LOGL0(op_data->op_name, op_data->op_namelen, tmp);

	}
}

void mdc_close_pack(struct ptlrpc_request *req, struct md_op_data *op_data)
{
	struct mdt_ioepoch *epoch;
	struct mdt_rec_setattr *rec;

	epoch = req_capsule_client_get(&req->rq_pill, &RMF_MDT_EPOCH);
	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);

	mdc_setattr_pack_rec(rec, op_data);
	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
	mdc_ioepoch_pack(epoch, op_data);
}

static int mdc_req_avail(struct client_obd *cli, struct mdc_cache_waiter *mcw)
{
	int rc;

	client_obd_list_lock(&cli->cl_loi_list_lock);
	rc = list_empty(&mcw->mcw_entry);
	client_obd_list_unlock(&cli->cl_loi_list_lock);
	return rc;
};

/* We record requests in flight in cli->cl_r_in_flight here.
 * There is only one write rpc possible in mdc anyway. If this to change
 * in the future - the code may need to be revisited. */
int mdc_enter_request(struct client_obd *cli)
{
	int rc = 0;
	struct mdc_cache_waiter mcw;
	struct l_wait_info lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP, NULL);

	client_obd_list_lock(&cli->cl_loi_list_lock);
	if (cli->cl_r_in_flight >= cli->cl_max_rpcs_in_flight) {
		list_add_tail(&mcw.mcw_entry, &cli->cl_cache_waiters);
		init_waitqueue_head(&mcw.mcw_waitq);
		client_obd_list_unlock(&cli->cl_loi_list_lock);
		rc = l_wait_event(mcw.mcw_waitq, mdc_req_avail(cli, &mcw), &lwi);
		if (rc) {
			client_obd_list_lock(&cli->cl_loi_list_lock);
			if (list_empty(&mcw.mcw_entry))
				cli->cl_r_in_flight--;
			list_del_init(&mcw.mcw_entry);
			client_obd_list_unlock(&cli->cl_loi_list_lock);
		}
	} else {
		cli->cl_r_in_flight++;
		client_obd_list_unlock(&cli->cl_loi_list_lock);
	}
	return rc;
}

void mdc_exit_request(struct client_obd *cli)
{
	struct list_head *l, *tmp;
	struct mdc_cache_waiter *mcw;

	client_obd_list_lock(&cli->cl_loi_list_lock);
	cli->cl_r_in_flight--;
	list_for_each_safe(l, tmp, &cli->cl_cache_waiters) {
		if (cli->cl_r_in_flight >= cli->cl_max_rpcs_in_flight) {
			/* No free request slots anymore */
			break;
		}

		mcw = list_entry(l, struct mdc_cache_waiter, mcw_entry);
		list_del_init(&mcw->mcw_entry);
		cli->cl_r_in_flight++;
		wake_up(&mcw->mcw_waitq);
	}
	/* Empty waiting list? Decrease reqs in-flight number */

	client_obd_list_unlock(&cli->cl_loi_list_lock);
}
Commit	Line	Data
d7e09d03 PT	1	/*
	2	* GPL HEADER START
	3	*
	4	* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
	5	*
	6	* This program is free software; you can redistribute it and/or modify
	7	* it under the terms of the GNU General Public License version 2 only,
	8	* as published by the Free Software Foundation.
	9	*
	10	* This program is distributed in the hope that it will be useful, but
	11	* WITHOUT ANY WARRANTY; without even the implied warranty of
	12	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	13	* General Public License version 2 for more details (a copy is included
	14	* in the LICENSE file that accompanied this code).
	15	*
	16	* You should have received a copy of the GNU General Public License
	17	* version 2 along with this program; If not, see
	18	* http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
	19	*
	20	* Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
	21	* CA 95054 USA or visit www.sun.com if you need additional information or
	22	* have any questions.
	23	*
	24	* GPL HEADER END
	25	*/
	26	/*
	27	* Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
	28	* Use is subject to license terms.
	29	*
	30	* Copyright (c) 2011, 2012, Intel Corporation.
	31	*/
	32	/*
	33	* This file is part of Lustre, http://www.lustre.org/
	34	* Lustre is a trademark of Sun Microsystems, Inc.
	35	*/
	36
	37	#define DEBUG_SUBSYSTEM S_MDC
	38	#include <lustre_net.h>
	39	#include <lustre/lustre_idl.h>
	40	#include "mdc_internal.h"
	41
	42
	43	static void __mdc_pack_body(struct mdt_body *b, __u32 suppgid)
	44	{
	45	LASSERT (b != NULL);
	46
	47	b->suppgid = suppgid;
4b1a25f0 PT	48	b->uid = from_kuid(&init_user_ns, current_uid());
	49	b->gid = from_kgid(&init_user_ns, current_gid());
	50	b->fsuid = from_kuid(&init_user_ns, current_fsuid());
	51	b->fsgid = from_kgid(&init_user_ns, current_fsgid());
d7e09d03 PT	52	b->capability = cfs_curproc_cap_pack();
	53	}
	54
	55	void mdc_pack_capa(struct ptlrpc_request req, const struct req_msg_field field,
	56	struct obd_capa *oc)
	57	{
	58	struct req_capsule *pill = &req->rq_pill;
	59	struct lustre_capa *c;
	60
	61	if (oc == NULL) {
	62	LASSERT(req_capsule_get_size(pill, field, RCL_CLIENT) == 0);
	63	return;
	64	}
	65
	66	c = req_capsule_client_get(pill, field);
	67	LASSERT(c != NULL);
	68	capa_cpy(c, oc);
	69	DEBUG_CAPA(D_SEC, c, "pack");
	70	}
	71
	72	void mdc_is_subdir_pack(struct ptlrpc_request req, const struct lu_fid pfid,
	73	const struct lu_fid *cfid, int flags)
	74	{
	75	struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
	76	&RMF_MDT_BODY);
	77
	78	if (pfid) {
	79	b->fid1 = *pfid;
	80	b->valid = OBD_MD_FLID;
	81	}
	82	if (cfid)
	83	b->fid2 = *cfid;
	84	b->flags = flags;
	85	}
	86
	87	void mdc_swap_layouts_pack(struct ptlrpc_request *req,
	88	struct md_op_data *op_data)
	89	{
	90	struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
	91	&RMF_MDT_BODY);
	92
	93	__mdc_pack_body(b, op_data->op_suppgids[0]);
	94	b->fid1 = op_data->op_fid1;
	95	b->fid2 = op_data->op_fid2;
	96	b->valid \|= OBD_MD_FLID;
	97
	98	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
	99	mdc_pack_capa(req, &RMF_CAPA2, op_data->op_capa2);
	100	}
	101
	102	void mdc_pack_body(struct ptlrpc_request *req,
	103	const struct lu_fid fid, struct obd_capa oc,
	104	__u64 valid, int ea_size, __u32 suppgid, int flags)
	105	{
	106	struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
	107	&RMF_MDT_BODY);
	108	LASSERT(b != NULL);
	109	b->valid = valid;
	110	b->eadatasize = ea_size;
	111	b->flags = flags;
	112	__mdc_pack_body(b, suppgid);
	113	if (fid) {
	114	b->fid1 = *fid;
	115	b->valid \|= OBD_MD_FLID;
116	mdc_pack_capa(req, &RMF_CAPA1, oc);
117	}
118	}
119
120	void mdc_readdir_pack(struct ptlrpc_request *req, __u64 pgoff,
121	__u32 size, const struct lu_fid fid, struct obd_capa oc)
122	{
123	struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
124	&RMF_MDT_BODY);
125	b->fid1 = *fid;
126	b->valid \|= OBD_MD_FLID;
127	b->size = pgoff; /* !! */
128	b->nlink = size; /* !! */
129	__mdc_pack_body(b, -1);
130	b->mode = LUDA_FID \| LUDA_TYPE;
131
132	mdc_pack_capa(req, &RMF_CAPA1, oc);
133	}
134
135	/* packing of MDS records */
136	void mdc_create_pack(struct ptlrpc_request req, struct md_op_data op_data,
137	const void *data, int datalen, __u32 mode,
138	__u32 uid, __u32 gid, cfs_cap_t cap_effective, __u64 rdev)
139	{
140	struct mdt_rec_create *rec;
141	char *tmp;
142	__u64 flags;
143
144	CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_create));
145	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
146
147
148	rec->cr_opcode = REINT_CREATE;
149	rec->cr_fsuid = uid;
150	rec->cr_fsgid = gid;
151	rec->cr_cap = cap_effective;
152	rec->cr_fid1 = op_data->op_fid1;
153	rec->cr_fid2 = op_data->op_fid2;
154	rec->cr_mode = mode;
155	rec->cr_rdev = rdev;
156	rec->cr_time = op_data->op_mod_time;
157	rec->cr_suppgid1 = op_data->op_suppgids[0];
158	rec->cr_suppgid2 = op_data->op_suppgids[1];
159	flags = op_data->op_flags & MF_SOM_LOCAL_FLAGS;
160	if (op_data->op_bias & MDS_CREATE_VOLATILE)
161	flags \|= MDS_OPEN_VOLATILE;
162	set_mrc_cr_flags(rec, flags);
163	rec->cr_bias = op_data->op_bias;
164	rec->cr_umask = current_umask();
165
166	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
167
168	tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
169	LOGL0(op_data->op_name, op_data->op_namelen, tmp);
170
171	if (data) {
172	tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
173	memcpy(tmp, data, datalen);
174	}
175	}
176
177	static __u64 mds_pack_open_flags(__u32 flags, __u32 mode)
178	{
179	__u64 cr_flags = (flags & (FMODE_READ \| FMODE_WRITE \|
180	MDS_OPEN_HAS_EA \| MDS_OPEN_HAS_OBJS \|
181	MDS_OPEN_OWNEROVERRIDE \| MDS_OPEN_LOCK \|
182	MDS_OPEN_BY_FID));
183	if (flags & O_CREAT)
184	cr_flags \|= MDS_OPEN_CREAT;
185	if (flags & O_EXCL)
186	cr_flags \|= MDS_OPEN_EXCL;
187	if (flags & O_TRUNC)
188	cr_flags \|= MDS_OPEN_TRUNC;
189	if (flags & O_APPEND)
190	cr_flags \|= MDS_OPEN_APPEND;
191	if (flags & O_SYNC)
192	cr_flags \|= MDS_OPEN_SYNC;
193	if (flags & O_DIRECTORY)
194	cr_flags \|= MDS_OPEN_DIRECTORY;
195	#ifdef FMODE_EXEC
196	if (flags & FMODE_EXEC)
197	cr_flags \|= MDS_FMODE_EXEC;
198	#endif
199	if (flags & O_LOV_DELAY_CREATE)
200	cr_flags \|= MDS_OPEN_DELAY_CREATE;
201
7d5ed06b	202	if (flags & O_NONBLOCK)
d7e09d03 PT	203	cr_flags \|= MDS_OPEN_NORESTORE;
	204
	205	return cr_flags;
	206	}
	207
	208	/* packing of MDS records */
	209	void mdc_open_pack(struct ptlrpc_request req, struct md_op_data op_data,
	210	__u32 mode, __u64 rdev, __u32 flags, const void *lmm,
	211	int lmmlen)
	212	{
	213	struct mdt_rec_create *rec;
	214	char *tmp;
	215	__u64 cr_flags;
	216
	217	CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_create));
	218	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
	219
	220	/* XXX do something about time, uid, gid */
	221	rec->cr_opcode = REINT_OPEN;
4b1a25f0 PT	222	rec->cr_fsuid = from_kuid(&init_user_ns, current_fsuid());
4b1a25f0 PT	223	rec->cr_fsgid = from_kgid(&init_user_ns, current_fsgid());
d7e09d03 PT	224	rec->cr_cap = cfs_curproc_cap_pack();
	225	if (op_data != NULL) {
	226	rec->cr_fid1 = op_data->op_fid1;
	227	rec->cr_fid2 = op_data->op_fid2;
	228	}
	229	rec->cr_mode = mode;
	230	cr_flags = mds_pack_open_flags(flags, mode);
	231	rec->cr_rdev = rdev;
	232	rec->cr_time = op_data->op_mod_time;
	233	rec->cr_suppgid1 = op_data->op_suppgids[0];
	234	rec->cr_suppgid2 = op_data->op_suppgids[1];
	235	rec->cr_bias = op_data->op_bias;
	236	rec->cr_umask = current_umask();
	237
	238	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
	239	/* the next buffer is child capa, which is used for replay,
	240	* will be packed from the data in reply message. */
	241
	242	if (op_data->op_name) {
	243	tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
	244	LOGL0(op_data->op_name, op_data->op_namelen, tmp);
	245	if (op_data->op_bias & MDS_CREATE_VOLATILE)
	246	cr_flags \|= MDS_OPEN_VOLATILE;
	247	}
	248
	249	if (lmm) {
	250	cr_flags \|= MDS_OPEN_HAS_EA;
	251	tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
	252	memcpy(tmp, lmm, lmmlen);
	253	}
	254	set_mrc_cr_flags(rec, cr_flags);
	255	}
	256
	257	static inline __u64 attr_pack(unsigned int ia_valid) {
	258	__u64 sa_valid = 0;
	259
	260	if (ia_valid & ATTR_MODE)
	261	sa_valid \|= MDS_ATTR_MODE;
	262	if (ia_valid & ATTR_UID)
	263	sa_valid \|= MDS_ATTR_UID;
	264	if (ia_valid & ATTR_GID)
	265	sa_valid \|= MDS_ATTR_GID;
	266	if (ia_valid & ATTR_SIZE)
	267	sa_valid \|= MDS_ATTR_SIZE;
	268	if (ia_valid & ATTR_ATIME)
	269	sa_valid \|= MDS_ATTR_ATIME;
	270	if (ia_valid & ATTR_MTIME)
	271	sa_valid \|= MDS_ATTR_MTIME;
	272	if (ia_valid & ATTR_CTIME)
	273	sa_valid \|= MDS_ATTR_CTIME;
	274	if (ia_valid & ATTR_ATIME_SET)
	275	sa_valid \|= MDS_ATTR_ATIME_SET;
	276	if (ia_valid & ATTR_MTIME_SET)
	277	sa_valid \|= MDS_ATTR_MTIME_SET;
	278	if (ia_valid & ATTR_FORCE)
	279	sa_valid \|= MDS_ATTR_FORCE;
	280	if (ia_valid & ATTR_ATTR_FLAG)
	281	sa_valid \|= MDS_ATTR_ATTR_FLAG;
	282	if (ia_valid & ATTR_KILL_SUID)
	283	sa_valid \|= MDS_ATTR_KILL_SUID;
	284	if (ia_valid & ATTR_KILL_SGID)
	285	sa_valid \|= MDS_ATTR_KILL_SGID;
	286	if (ia_valid & ATTR_CTIME_SET)
	287	sa_valid \|= MDS_ATTR_CTIME_SET;
288	if (ia_valid & ATTR_FROM_OPEN)
289	sa_valid \|= MDS_ATTR_FROM_OPEN;
290	if (ia_valid & ATTR_BLOCKS)
291	sa_valid \|= MDS_ATTR_BLOCKS;
292	if (ia_valid & MDS_OPEN_OWNEROVERRIDE)
293	/* NFSD hack (see bug 5781) */
294	sa_valid \|= MDS_OPEN_OWNEROVERRIDE;
295	return sa_valid;
296	}
297
298	static void mdc_setattr_pack_rec(struct mdt_rec_setattr *rec,
299	struct md_op_data *op_data)
300	{
301	rec->sa_opcode = REINT_SETATTR;
4b1a25f0 PT	302	rec->sa_fsuid = from_kuid(&init_user_ns, current_fsuid());
4b1a25f0 PT	303	rec->sa_fsgid = from_kgid(&init_user_ns, current_fsgid());
d7e09d03 PT	304	rec->sa_cap = cfs_curproc_cap_pack();
	305	rec->sa_suppgid = -1;
	306
	307	rec->sa_fid = op_data->op_fid1;
	308	rec->sa_valid = attr_pack(op_data->op_attr.ia_valid);
	309	rec->sa_mode = op_data->op_attr.ia_mode;
4b1a25f0 PT	310	rec->sa_uid = from_kuid(&init_user_ns, op_data->op_attr.ia_uid);
4b1a25f0 PT	311	rec->sa_gid = from_kgid(&init_user_ns, op_data->op_attr.ia_gid);
d7e09d03 PT	312	rec->sa_size = op_data->op_attr.ia_size;
	313	rec->sa_blocks = op_data->op_attr_blocks;
	314	rec->sa_atime = LTIME_S(op_data->op_attr.ia_atime);
	315	rec->sa_mtime = LTIME_S(op_data->op_attr.ia_mtime);
	316	rec->sa_ctime = LTIME_S(op_data->op_attr.ia_ctime);
	317	rec->sa_attr_flags = ((struct ll_iattr *)&op_data->op_attr)->ia_attr_flags;
	318	if ((op_data->op_attr.ia_valid & ATTR_GID) &&
4b1a25f0 PT	319	in_group_p(op_data->op_attr.ia_gid))
	320	rec->sa_suppgid =
	321	from_kgid(&init_user_ns, op_data->op_attr.ia_gid);
d7e09d03 PT	322	else
	323	rec->sa_suppgid = op_data->op_suppgids[0];
	324
	325	rec->sa_bias = op_data->op_bias;
	326	}
	327
	328	static void mdc_ioepoch_pack(struct mdt_ioepoch *epoch,
	329	struct md_op_data *op_data)
	330	{
	331	memcpy(&epoch->handle, &op_data->op_handle, sizeof(epoch->handle));
	332	epoch->ioepoch = op_data->op_ioepoch;
	333	epoch->flags = op_data->op_flags & MF_SOM_LOCAL_FLAGS;
	334	}
	335
	336	void mdc_setattr_pack(struct ptlrpc_request req, struct md_op_data op_data,
	337	void ea, int ealen, void ea2, int ea2len)
	338	{
	339	struct mdt_rec_setattr *rec;
	340	struct mdt_ioepoch *epoch;
	341	struct lov_user_md *lum = NULL;
	342
	343	CLASSERT(sizeof(struct mdt_rec_reint) ==sizeof(struct mdt_rec_setattr));
	344	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
	345	mdc_setattr_pack_rec(rec, op_data);
	346
	347	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
	348
	349	if (op_data->op_flags & (MF_SOM_CHANGE \| MF_EPOCH_OPEN)) {
	350	epoch = req_capsule_client_get(&req->rq_pill, &RMF_MDT_EPOCH);
	351	mdc_ioepoch_pack(epoch, op_data);
	352	}
	353
	354	if (ealen == 0)
	355	return;
	356
	357	lum = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
	358	if (ea == NULL) { /* Remove LOV EA */
	359	lum->lmm_magic = LOV_USER_MAGIC_V1;
	360	lum->lmm_stripe_size = 0;
	361	lum->lmm_stripe_count = 0;
	362	lum->lmm_stripe_offset = (typeof(lum->lmm_stripe_offset))(-1);
	363	} else {
	364	memcpy(lum, ea, ealen);
	365	}
	366
	367	if (ea2len == 0)
	368	return;
	369
	370	memcpy(req_capsule_client_get(&req->rq_pill, &RMF_LOGCOOKIES), ea2,
	371	ea2len);
	372	}
	373
	374	void mdc_unlink_pack(struct ptlrpc_request req, struct md_op_data op_data)
	375	{
	376	struct mdt_rec_unlink *rec;
	377	char *tmp;
	378
	379	CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_unlink));
	380	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
	381	LASSERT(rec != NULL);
	382
	383	rec->ul_opcode = op_data->op_cli_flags & CLI_RM_ENTRY ?
	384	REINT_RMENTRY : REINT_UNLINK;
	385	rec->ul_fsuid = op_data->op_fsuid;
386	rec->ul_fsgid = op_data->op_fsgid;
387	rec->ul_cap = op_data->op_cap;
388	rec->ul_mode = op_data->op_mode;
389	rec->ul_suppgid1= op_data->op_suppgids[0];
390	rec->ul_suppgid2= -1;
391	rec->ul_fid1 = op_data->op_fid1;
392	rec->ul_fid2 = op_data->op_fid2;
393	rec->ul_time = op_data->op_mod_time;
394	rec->ul_bias = op_data->op_bias;
395
396	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
397
398	tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
399	LASSERT(tmp != NULL);
400	LOGL0(op_data->op_name, op_data->op_namelen, tmp);
401	}
402
403	void mdc_link_pack(struct ptlrpc_request req, struct md_op_data op_data)
404	{
405	struct mdt_rec_link *rec;
406	char *tmp;
407
408	CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_link));
409	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
410	LASSERT (rec != NULL);
411
412	rec->lk_opcode = REINT_LINK;
413	rec->lk_fsuid = op_data->op_fsuid;//current->fsuid;
414	rec->lk_fsgid = op_data->op_fsgid;//current->fsgid;
415	rec->lk_cap = op_data->op_cap;//current->cap_effective;
416	rec->lk_suppgid1 = op_data->op_suppgids[0];
417	rec->lk_suppgid2 = op_data->op_suppgids[1];
418	rec->lk_fid1 = op_data->op_fid1;
419	rec->lk_fid2 = op_data->op_fid2;
420	rec->lk_time = op_data->op_mod_time;
421	rec->lk_bias = op_data->op_bias;
422
423	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
424	mdc_pack_capa(req, &RMF_CAPA2, op_data->op_capa2);
425
426	tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
427	LOGL0(op_data->op_name, op_data->op_namelen, tmp);
428	}
429
430	void mdc_rename_pack(struct ptlrpc_request req, struct md_op_data op_data,
431	const char old, int oldlen, const char new, int newlen)
432	{
433	struct mdt_rec_rename *rec;
434	char *tmp;
435
436	CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_rename));
437	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
438
439	/* XXX do something about time, uid, gid */
440	rec->rn_opcode = REINT_RENAME;
441	rec->rn_fsuid = op_data->op_fsuid;
442	rec->rn_fsgid = op_data->op_fsgid;
443	rec->rn_cap = op_data->op_cap;
444	rec->rn_suppgid1 = op_data->op_suppgids[0];
445	rec->rn_suppgid2 = op_data->op_suppgids[1];
446	rec->rn_fid1 = op_data->op_fid1;
447	rec->rn_fid2 = op_data->op_fid2;
448	rec->rn_time = op_data->op_mod_time;
449	rec->rn_mode = op_data->op_mode;
450	rec->rn_bias = op_data->op_bias;
451
452	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
453	mdc_pack_capa(req, &RMF_CAPA2, op_data->op_capa2);
454
455	tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
456	LOGL0(old, oldlen, tmp);
457
458	if (new) {
459	tmp = req_capsule_client_get(&req->rq_pill, &RMF_SYMTGT);
460	LOGL0(new, newlen, tmp);
461	}
462	}
463
464	void mdc_getattr_pack(struct ptlrpc_request *req, __u64 valid, int flags,
465	struct md_op_data *op_data, int ea_size)
466	{
467	struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
468	&RMF_MDT_BODY);
469
470	b->valid = valid;
471	if (op_data->op_bias & MDS_CHECK_SPLIT)
472	b->valid \|= OBD_MD_FLCKSPLIT;
473	if (op_data->op_bias & MDS_CROSS_REF)
474	b->valid \|= OBD_MD_FLCROSSREF;
475	b->eadatasize = ea_size;
476	b->flags = flags;
477	__mdc_pack_body(b, op_data->op_suppgids[0]);
478
479	b->fid1 = op_data->op_fid1;
480	b->fid2 = op_data->op_fid2;
481	b->valid \|= OBD_MD_FLID;
482
483	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
484
485	if (op_data->op_name) {
486	char *tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
487	LOGL0(op_data->op_name, op_data->op_namelen, tmp);
488
489	}
490	}
491
492	void mdc_close_pack(struct ptlrpc_request req, struct md_op_data op_data)
493	{
494	struct mdt_ioepoch *epoch;
495	struct mdt_rec_setattr *rec;
496
497	epoch = req_capsule_client_get(&req->rq_pill, &RMF_MDT_EPOCH);
498	rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
499
500	mdc_setattr_pack_rec(rec, op_data);
501	mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1);
502	mdc_ioepoch_pack(epoch, op_data);
503	}
504
505	static int mdc_req_avail(struct client_obd cli, struct mdc_cache_waiter mcw)
506	{
507	int rc;
29aaf496	508
d7e09d03 PT	509	client_obd_list_lock(&cli->cl_loi_list_lock);
	510	rc = list_empty(&mcw->mcw_entry);
	511	client_obd_list_unlock(&cli->cl_loi_list_lock);
0a3bdb00	512	return rc;
d7e09d03 PT	513	};
	514
	515	/* We record requests in flight in cli->cl_r_in_flight here.
	516	* There is only one write rpc possible in mdc anyway. If this to change
	517	* in the future - the code may need to be revisited. */
	518	int mdc_enter_request(struct client_obd *cli)
	519	{
	520	int rc = 0;
	521	struct mdc_cache_waiter mcw;
	522	struct l_wait_info lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP, NULL);
	523
	524	client_obd_list_lock(&cli->cl_loi_list_lock);
	525	if (cli->cl_r_in_flight >= cli->cl_max_rpcs_in_flight) {
	526	list_add_tail(&mcw.mcw_entry, &cli->cl_cache_waiters);
	527	init_waitqueue_head(&mcw.mcw_waitq);
	528	client_obd_list_unlock(&cli->cl_loi_list_lock);
	529	rc = l_wait_event(mcw.mcw_waitq, mdc_req_avail(cli, &mcw), &lwi);
	530	if (rc) {
	531	client_obd_list_lock(&cli->cl_loi_list_lock);
	532	if (list_empty(&mcw.mcw_entry))
	533	cli->cl_r_in_flight--;
	534	list_del_init(&mcw.mcw_entry);
	535	client_obd_list_unlock(&cli->cl_loi_list_lock);
	536	}
	537	} else {
	538	cli->cl_r_in_flight++;
	539	client_obd_list_unlock(&cli->cl_loi_list_lock);
	540	}
	541	return rc;
	542	}
	543
	544	void mdc_exit_request(struct client_obd *cli)
	545	{
	546	struct list_head l, tmp;
	547	struct mdc_cache_waiter *mcw;
	548
	549	client_obd_list_lock(&cli->cl_loi_list_lock);
	550	cli->cl_r_in_flight--;
	551	list_for_each_safe(l, tmp, &cli->cl_cache_waiters) {
	552	if (cli->cl_r_in_flight >= cli->cl_max_rpcs_in_flight) {
	553	/* No free request slots anymore */
	554	break;
	555	}
	556
	557	mcw = list_entry(l, struct mdc_cache_waiter, mcw_entry);
	558	list_del_init(&mcw->mcw_entry);
	559	cli->cl_r_in_flight++;
	560	wake_up(&mcw->mcw_waitq);
	561	}
	562	/* Empty waiting list? Decrease reqs in-flight number */
	563
	564	client_obd_list_unlock(&cli->cl_loi_list_lock);
	565	}