[linux-2.6-block.git] / fs / nfs / filelayout / filelayoutdev.c

/*
 *  Device operations for the pnfs nfs4 file layout driver.
 *
 *  Copyright (c) 2002
 *  The Regents of the University of Michigan
 *  All Rights Reserved
 *
 *  Dean Hildebrand <dhildebz@umich.edu>
 *  Garth Goodson   <Garth.Goodson@netapp.com>
 *
 *  Permission is granted to use, copy, create derivative works, and
 *  redistribute this software and such derivative works for any purpose,
 *  so long as the name of the University of Michigan is not used in
 *  any advertising or publicity pertaining to the use or distribution
 *  of this software without specific, written prior authorization. If
 *  the above copyright notice or any other identification of the
 *  University of Michigan is included in any copy of any portion of
 *  this software, then the disclaimer below must also be included.
 *
 *  This software is provided as is, without representation or warranty
 *  of any kind either express or implied, including without limitation
 *  the implied warranties of merchantability, fitness for a particular
 *  purpose, or noninfringement.  The Regents of the University of
 *  Michigan shall not be liable for any damages, including special,
 *  indirect, incidental, or consequential damages, with respect to any
 *  claim arising out of or in connection with the use of the software,
 *  even if it has been or is hereafter advised of the possibility of
 *  such damages.
 */

#include <linux/nfs_fs.h>
#include <linux/vmalloc.h>
#include <linux/module.h>
#include <linux/sunrpc/addr.h>

#include "../internal.h"
#include "../nfs4session.h"
#include "filelayout.h"

#define NFSDBG_FACILITY		NFSDBG_PNFS_LD

static unsigned int dataserver_timeo = NFS4_DEF_DS_TIMEO;
static unsigned int dataserver_retrans = NFS4_DEF_DS_RETRANS;

/*
 * Create an rpc connection to the nfs4_pnfs_ds data server
 * Currently only supports IPv4 and IPv6 addresses
 */
static int
nfs4_ds_connect(struct nfs_server *mds_srv, struct nfs4_pnfs_ds *ds)
{
	struct nfs_client *clp = ERR_PTR(-EIO);
	struct nfs4_pnfs_ds_addr *da;
	int status = 0;

	dprintk("--> %s DS %s au_flavor %d\n", __func__, ds->ds_remotestr,
		mds_srv->nfs_client->cl_rpcclient->cl_auth->au_flavor);

	list_for_each_entry(da, &ds->ds_addrs, da_node) {
		dprintk("%s: DS %s: trying address %s\n",
			__func__, ds->ds_remotestr, da->da_remotestr);

		clp = nfs4_set_ds_client(mds_srv->nfs_client,
					(struct sockaddr *)&da->da_addr,
					da->da_addrlen, IPPROTO_TCP,
					dataserver_timeo, dataserver_retrans);
		if (!IS_ERR(clp))
			break;
	}

	if (IS_ERR(clp)) {
		status = PTR_ERR(clp);
		goto out;
	}

	status = nfs4_init_ds_session(clp, mds_srv->nfs_client->cl_lease_time);
	if (status)
		goto out_put;

	smp_wmb();
	ds->ds_clp = clp;
	dprintk("%s [new] addr: %s\n", __func__, ds->ds_remotestr);
out:
	return status;
out_put:
	nfs_put_client(clp);
	goto out;
}

void
nfs4_fl_free_deviceid(struct nfs4_file_layout_dsaddr *dsaddr)
{
	struct nfs4_pnfs_ds *ds;
	int i;

	nfs4_print_deviceid(&dsaddr->id_node.deviceid);

	for (i = 0; i < dsaddr->ds_num; i++) {
		ds = dsaddr->ds_list[i];
		if (ds != NULL)
			nfs4_pnfs_ds_put(ds);
	}
	kfree(dsaddr->stripe_indices);
	kfree(dsaddr);
}

/*
 * Currently only supports ipv4, ipv6 and one multi-path address.
 */
static struct nfs4_pnfs_ds_addr *
decode_ds_addr(struct net *net, struct xdr_stream *streamp, gfp_t gfp_flags)
{
	struct nfs4_pnfs_ds_addr *da = NULL;
	char *buf, *portstr;
	__be16 port;
	int nlen, rlen;
	int tmp[2];
	__be32 *p;
	char *netid, *match_netid;
	size_t len, match_netid_len;
	char *startsep = "";
	char *endsep = "";


	/* r_netid */
	p = xdr_inline_decode(streamp, 4);
	if (unlikely(!p))
		goto out_err;
	nlen = be32_to_cpup(p++);

	p = xdr_inline_decode(streamp, nlen);
	if (unlikely(!p))
		goto out_err;

	netid = kmalloc(nlen+1, gfp_flags);
	if (unlikely(!netid))
		goto out_err;

	netid[nlen] = '\0';
	memcpy(netid, p, nlen);

	/* r_addr: ip/ip6addr with port in dec octets - see RFC 5665 */
	p = xdr_inline_decode(streamp, 4);
	if (unlikely(!p))
		goto out_free_netid;
	rlen = be32_to_cpup(p);

	p = xdr_inline_decode(streamp, rlen);
	if (unlikely(!p))
		goto out_free_netid;

	/* port is ".ABC.DEF", 8 chars max */
	if (rlen > INET6_ADDRSTRLEN + IPV6_SCOPE_ID_LEN + 8) {
		dprintk("%s: Invalid address, length %d\n", __func__,
			rlen);
		goto out_free_netid;
	}
	buf = kmalloc(rlen + 1, gfp_flags);
	if (!buf) {
		dprintk("%s: Not enough memory\n", __func__);
		goto out_free_netid;
	}
	buf[rlen] = '\0';
	memcpy(buf, p, rlen);

	/* replace port '.' with '-' */
	portstr = strrchr(buf, '.');
	if (!portstr) {
		dprintk("%s: Failed finding expected dot in port\n",
			__func__);
		goto out_free_buf;
	}
	*portstr = '-';

	/* find '.' between address and port */
	portstr = strrchr(buf, '.');
	if (!portstr) {
		dprintk("%s: Failed finding expected dot between address and "
			"port\n", __func__);
		goto out_free_buf;
	}
	*portstr = '\0';

	da = kzalloc(sizeof(*da), gfp_flags);
	if (unlikely(!da))
		goto out_free_buf;

	INIT_LIST_HEAD(&da->da_node);

	if (!rpc_pton(net, buf, portstr-buf, (struct sockaddr *)&da->da_addr,
		      sizeof(da->da_addr))) {
		dprintk("%s: error parsing address %s\n", __func__, buf);
		goto out_free_da;
	}

	portstr++;
	sscanf(portstr, "%d-%d", &tmp[0], &tmp[1]);
	port = htons((tmp[0] << 8) | (tmp[1]));

	switch (da->da_addr.ss_family) {
	case AF_INET:
		((struct sockaddr_in *)&da->da_addr)->sin_port = port;
		da->da_addrlen = sizeof(struct sockaddr_in);
		match_netid = "tcp";
		match_netid_len = 3;
		break;

	case AF_INET6:
		((struct sockaddr_in6 *)&da->da_addr)->sin6_port = port;
		da->da_addrlen = sizeof(struct sockaddr_in6);
		match_netid = "tcp6";
		match_netid_len = 4;
		startsep = "[";
		endsep = "]";
		break;

	default:
		dprintk("%s: unsupported address family: %u\n",
			__func__, da->da_addr.ss_family);
		goto out_free_da;
	}

	if (nlen != match_netid_len || strncmp(netid, match_netid, nlen)) {
		dprintk("%s: ERROR: r_netid \"%s\" != \"%s\"\n",
			__func__, netid, match_netid);
		goto out_free_da;
	}

	/* save human readable address */
	len = strlen(startsep) + strlen(buf) + strlen(endsep) + 7;
	da->da_remotestr = kzalloc(len, gfp_flags);

	/* NULL is ok, only used for dprintk */
	if (da->da_remotestr)
		snprintf(da->da_remotestr, len, "%s%s%s:%u", startsep,
			 buf, endsep, ntohs(port));

	dprintk("%s: Parsed DS addr %s\n", __func__, da->da_remotestr);
	kfree(buf);
	kfree(netid);
	return da;

out_free_da:
	kfree(da);
out_free_buf:
	dprintk("%s: Error parsing DS addr: %s\n", __func__, buf);
	kfree(buf);
out_free_netid:
	kfree(netid);
out_err:
	return NULL;
}

/* Decode opaque device data and return the result */
struct nfs4_file_layout_dsaddr *
nfs4_fl_alloc_deviceid_node(struct nfs_server *server, struct pnfs_device *pdev,
		gfp_t gfp_flags)
{
	int i;
	u32 cnt, num;
	u8 *indexp;
	__be32 *p;
	u8 *stripe_indices;
	u8 max_stripe_index;
	struct nfs4_file_layout_dsaddr *dsaddr = NULL;
	struct xdr_stream stream;
	struct xdr_buf buf;
	struct page *scratch;
	struct list_head dsaddrs;
	struct nfs4_pnfs_ds_addr *da;

	/* set up xdr stream */
	scratch = alloc_page(gfp_flags);
	if (!scratch)
		goto out_err;

	xdr_init_decode_pages(&stream, &buf, pdev->pages, pdev->pglen);
	xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE);

	/* Get the stripe count (number of stripe index) */
	p = xdr_inline_decode(&stream, 4);
	if (unlikely(!p))
		goto out_err_free_scratch;

	cnt = be32_to_cpup(p);
	dprintk("%s stripe count  %d\n", __func__, cnt);
	if (cnt > NFS4_PNFS_MAX_STRIPE_CNT) {
		printk(KERN_WARNING "NFS: %s: stripe count %d greater than "
		       "supported maximum %d\n", __func__,
			cnt, NFS4_PNFS_MAX_STRIPE_CNT);
		goto out_err_free_scratch;
	}

	/* read stripe indices */
	stripe_indices = kcalloc(cnt, sizeof(u8), gfp_flags);
	if (!stripe_indices)
		goto out_err_free_scratch;

	p = xdr_inline_decode(&stream, cnt << 2);
	if (unlikely(!p))
		goto out_err_free_stripe_indices;

	indexp = &stripe_indices[0];
	max_stripe_index = 0;
	for (i = 0; i < cnt; i++) {
		*indexp = be32_to_cpup(p++);
		max_stripe_index = max(max_stripe_index, *indexp);
		indexp++;
	}

	/* Check the multipath list count */
	p = xdr_inline_decode(&stream, 4);
	if (unlikely(!p))
		goto out_err_free_stripe_indices;

	num = be32_to_cpup(p);
	dprintk("%s ds_num %u\n", __func__, num);
	if (num > NFS4_PNFS_MAX_MULTI_CNT) {
		printk(KERN_WARNING "NFS: %s: multipath count %d greater than "
			"supported maximum %d\n", __func__,
			num, NFS4_PNFS_MAX_MULTI_CNT);
		goto out_err_free_stripe_indices;
	}

	/* validate stripe indices are all < num */
	if (max_stripe_index >= num) {
		printk(KERN_WARNING "NFS: %s: stripe index %u >= num ds %u\n",
			__func__, max_stripe_index, num);
		goto out_err_free_stripe_indices;
	}

	dsaddr = kzalloc(sizeof(*dsaddr) +
			(sizeof(struct nfs4_pnfs_ds *) * (num - 1)),
			gfp_flags);
	if (!dsaddr)
		goto out_err_free_stripe_indices;

	dsaddr->stripe_count = cnt;
	dsaddr->stripe_indices = stripe_indices;
	stripe_indices = NULL;
	dsaddr->ds_num = num;
	nfs4_init_deviceid_node(&dsaddr->id_node, server, &pdev->dev_id);

	INIT_LIST_HEAD(&dsaddrs);

	for (i = 0; i < dsaddr->ds_num; i++) {
		int j;
		u32 mp_count;

		p = xdr_inline_decode(&stream, 4);
		if (unlikely(!p))
			goto out_err_free_deviceid;

		mp_count = be32_to_cpup(p); /* multipath count */
		for (j = 0; j < mp_count; j++) {
			da = decode_ds_addr(server->nfs_client->cl_net,
					    &stream, gfp_flags);
			if (da)
				list_add_tail(&da->da_node, &dsaddrs);
		}
		if (list_empty(&dsaddrs)) {
			dprintk("%s: no suitable DS addresses found\n",
				__func__);
			goto out_err_free_deviceid;
		}

		dsaddr->ds_list[i] = nfs4_pnfs_ds_add(&dsaddrs, gfp_flags);
		if (!dsaddr->ds_list[i])
			goto out_err_drain_dsaddrs;

		/* If DS was already in cache, free ds addrs */
		while (!list_empty(&dsaddrs)) {
			da = list_first_entry(&dsaddrs,
					      struct nfs4_pnfs_ds_addr,
					      da_node);
			list_del_init(&da->da_node);
			kfree(da->da_remotestr);
			kfree(da);
		}
	}

	__free_page(scratch);
	return dsaddr;

out_err_drain_dsaddrs:
	while (!list_empty(&dsaddrs)) {
		da = list_first_entry(&dsaddrs, struct nfs4_pnfs_ds_addr,
				      da_node);
		list_del_init(&da->da_node);
		kfree(da->da_remotestr);
		kfree(da);
	}
out_err_free_deviceid:
	nfs4_fl_free_deviceid(dsaddr);
	/* stripe_indicies was part of dsaddr */
	goto out_err_free_scratch;
out_err_free_stripe_indices:
	kfree(stripe_indices);
out_err_free_scratch:
	__free_page(scratch);
out_err:
	dprintk("%s ERROR: returning NULL\n", __func__);
	return NULL;
}

void
nfs4_fl_put_deviceid(struct nfs4_file_layout_dsaddr *dsaddr)
{
	nfs4_put_deviceid_node(&dsaddr->id_node);
}

/*
 * Want res = (offset - layout->pattern_offset)/ layout->stripe_unit
 * Then: ((res + fsi) % dsaddr->stripe_count)
 */
u32
nfs4_fl_calc_j_index(struct pnfs_layout_segment *lseg, loff_t offset)
{
	struct nfs4_filelayout_segment *flseg = FILELAYOUT_LSEG(lseg);
	u64 tmp;

	tmp = offset - flseg->pattern_offset;
	do_div(tmp, flseg->stripe_unit);
	tmp += flseg->first_stripe_index;
	return do_div(tmp, flseg->dsaddr->stripe_count);
}

u32
nfs4_fl_calc_ds_index(struct pnfs_layout_segment *lseg, u32 j)
{
	return FILELAYOUT_LSEG(lseg)->dsaddr->stripe_indices[j];
}

struct nfs_fh *
nfs4_fl_select_ds_fh(struct pnfs_layout_segment *lseg, u32 j)
{
	struct nfs4_filelayout_segment *flseg = FILELAYOUT_LSEG(lseg);
	u32 i;

	if (flseg->stripe_type == STRIPE_SPARSE) {
		if (flseg->num_fh == 1)
			i = 0;
		else if (flseg->num_fh == 0)
			/* Use the MDS OPEN fh set in nfs_read_rpcsetup */
			return NULL;
		else
			i = nfs4_fl_calc_ds_index(lseg, j);
	} else
		i = j;
	return flseg->fh_array[i];
}

static void nfs4_wait_ds_connect(struct nfs4_pnfs_ds *ds)
{
	might_sleep();
	wait_on_bit_action(&ds->ds_state, NFS4DS_CONNECTING,
			   nfs_wait_bit_killable, TASK_KILLABLE);
}

static void nfs4_clear_ds_conn_bit(struct nfs4_pnfs_ds *ds)
{
	smp_mb__before_atomic();
	clear_bit(NFS4DS_CONNECTING, &ds->ds_state);
	smp_mb__after_atomic();
	wake_up_bit(&ds->ds_state, NFS4DS_CONNECTING);
}


struct nfs4_pnfs_ds *
nfs4_fl_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx)
{
	struct nfs4_file_layout_dsaddr *dsaddr = FILELAYOUT_LSEG(lseg)->dsaddr;
	struct nfs4_pnfs_ds *ds = dsaddr->ds_list[ds_idx];
	struct nfs4_deviceid_node *devid = FILELAYOUT_DEVID_NODE(lseg);
	struct nfs4_pnfs_ds *ret = ds;

	if (ds == NULL) {
		printk(KERN_ERR "NFS: %s: No data server for offset index %d\n",
			__func__, ds_idx);
		pnfs_generic_mark_devid_invalid(devid);
		goto out;
	}
	smp_rmb();
	if (ds->ds_clp)
		goto out_test_devid;

	if (test_and_set_bit(NFS4DS_CONNECTING, &ds->ds_state) == 0) {
		struct nfs_server *s = NFS_SERVER(lseg->pls_layout->plh_inode);
		int err;

		err = nfs4_ds_connect(s, ds);
		if (err)
			nfs4_mark_deviceid_unavailable(devid);
		nfs4_clear_ds_conn_bit(ds);
	} else {
		/* Either ds is connected, or ds is NULL */
		nfs4_wait_ds_connect(ds);
	}
out_test_devid:
	if (filelayout_test_devid_unavailable(devid))
		ret = NULL;
out:
	return ret;
}

module_param(dataserver_retrans, uint, 0644);
MODULE_PARM_DESC(dataserver_retrans, "The  number of times the NFSv4.1 client "
			"retries a request before it attempts further "
			" recovery  action.");
module_param(dataserver_timeo, uint, 0644);
MODULE_PARM_DESC(dataserver_timeo, "The time (in tenths of a second) the "
			"NFSv4.1  client  waits for a response from a "
			" data server before it retries an NFS request.");
Commit	Line	Data
16b374ca AA	1	/*
	2	* Device operations for the pnfs nfs4 file layout driver.
	3	*
	4	* Copyright (c) 2002
	5	* The Regents of the University of Michigan
	6	* All Rights Reserved
	7	*
	8	* Dean Hildebrand <dhildebz@umich.edu>
	9	* Garth Goodson <Garth.Goodson@netapp.com>
	10	*
	11	* Permission is granted to use, copy, create derivative works, and
	12	* redistribute this software and such derivative works for any purpose,
	13	* so long as the name of the University of Michigan is not used in
	14	* any advertising or publicity pertaining to the use or distribution
	15	* of this software without specific, written prior authorization. If
	16	* the above copyright notice or any other identification of the
	17	* University of Michigan is included in any copy of any portion of
	18	* this software, then the disclaimer below must also be included.
	19	*
	20	* This software is provided as is, without representation or warranty
	21	* of any kind either express or implied, including without limitation
	22	* the implied warranties of merchantability, fitness for a particular
	23	* purpose, or noninfringement. The Regents of the University of
	24	* Michigan shall not be liable for any damages, including special,
	25	* indirect, incidental, or consequential damages, with respect to any
	26	* claim arising out of or in connection with the use of the software,
	27	* even if it has been or is hereafter advised of the possibility of
	28	* such damages.
	29	*/
	30
	31	#include <linux/nfs_fs.h>
	32	#include <linux/vmalloc.h>
98fc685a	33	#include <linux/module.h>
5976687a	34	#include <linux/sunrpc/addr.h>
16b374ca	35
b5968725 TH	36	#include "../internal.h"
	37	#include "../nfs4session.h"
	38	#include "filelayout.h"
16b374ca AA	39
	40	#define NFSDBG_FACILITY NFSDBG_PNFS_LD
	41
98fc685a AA	42	static unsigned int dataserver_timeo = NFS4_DEF_DS_TIMEO;
	43	static unsigned int dataserver_retrans = NFS4_DEF_DS_RETRANS;
	44
d83217c1 AA	45	/*
d83217c1 AA	46	* Create an rpc connection to the nfs4_pnfs_ds data server
35dbbc99	47	* Currently only supports IPv4 and IPv6 addresses
d83217c1 AA	48	*/
	49	static int
	50	nfs4_ds_connect(struct nfs_server mds_srv, struct nfs4_pnfs_ds ds)
	51	{
7e574f0d	52	struct nfs_client *clp = ERR_PTR(-EIO);
14f9a607	53	struct nfs4_pnfs_ds_addr *da;
d83217c1 AA	54	int status = 0;
d83217c1 AA	55
14f9a607	56	dprintk("--> %s DS %s au_flavor %d\n", __func__, ds->ds_remotestr,
d83217c1 AA	57	mds_srv->nfs_client->cl_rpcclient->cl_auth->au_flavor);
d83217c1 AA	58
7e574f0d WAA	59	list_for_each_entry(da, &ds->ds_addrs, da_node) {
	60	dprintk("%s: DS %s: trying address %s\n",
	61	__func__, ds->ds_remotestr, da->da_remotestr);
14f9a607	62
7e574f0d	63	clp = nfs4_set_ds_client(mds_srv->nfs_client,
98fc685a AA	64	(struct sockaddr *)&da->da_addr,
	65	da->da_addrlen, IPPROTO_TCP,
	66	dataserver_timeo, dataserver_retrans);
7e574f0d WAA	67	if (!IS_ERR(clp))
	68	break;
	69	}
	70
d83217c1 AA	71	if (IS_ERR(clp)) {
	72	status = PTR_ERR(clp);
	73	goto out;
	74	}
	75
7b38c368	76	status = nfs4_init_ds_session(clp, mds_srv->nfs_client->cl_lease_time);
d83217c1 AA	77	if (status)
	78	goto out_put;
	79
acd65e5b	80	smp_wmb();
d83217c1	81	ds->ds_clp = clp;
c9895cb6	82	dprintk("%s [new] addr: %s\n", __func__, ds->ds_remotestr);
d83217c1 AA	83	out:
	84	return status;
	85	out_put:
	86	nfs_put_client(clp);
	87	goto out;
	88	}
	89
1775bc34	90	void
16b374ca AA	91	nfs4_fl_free_deviceid(struct nfs4_file_layout_dsaddr *dsaddr)
	92	{
	93	struct nfs4_pnfs_ds *ds;
	94	int i;
	95
a1eaecbc	96	nfs4_print_deviceid(&dsaddr->id_node.deviceid);
16b374ca AA	97
	98	for (i = 0; i < dsaddr->ds_num; i++) {
	99	ds = dsaddr->ds_list[i];
875ae069 PT	100	if (ds != NULL)
875ae069 PT	101	nfs4_pnfs_ds_put(ds);
16b374ca AA	102	}
	103	kfree(dsaddr->stripe_indices);
	104	kfree(dsaddr);
	105	}
	106
16b374ca	107	/*
c9895cb6	108	* Currently only supports ipv4, ipv6 and one multi-path address.
16b374ca	109	*/
14f9a607	110	static struct nfs4_pnfs_ds_addr *
17094272	111	decode_ds_addr(struct net net, struct xdr_stream streamp, gfp_t gfp_flags)
16b374ca	112	{
14f9a607	113	struct nfs4_pnfs_ds_addr *da = NULL;
c9895cb6	114	char buf, portstr;
13fff2f3	115	__be16 port;
c9895cb6	116	int nlen, rlen;
16b374ca	117	int tmp[2];
35124a09	118	__be32 *p;
c9895cb6	119	char netid, match_netid;
14f9a607 WAA	120	size_t len, match_netid_len;
	121	char *startsep = "";
	122	char *endsep = "";
	123
16b374ca AA	124
16b374ca AA	125	/* r_netid */
35124a09 WAA	126	p = xdr_inline_decode(streamp, 4);
	127	if (unlikely(!p))
	128	goto out_err;
16b374ca	129	nlen = be32_to_cpup(p++);
16b374ca	130
35124a09 WAA	131	p = xdr_inline_decode(streamp, nlen);
	132	if (unlikely(!p))
	133	goto out_err;
16b374ca	134
c9895cb6 WAA	135	netid = kmalloc(nlen+1, gfp_flags);
c9895cb6 WAA	136	if (unlikely(!netid))
16b374ca	137	goto out_err;
16b374ca	138
c9895cb6 WAA	139	netid[nlen] = '\0';
	140	memcpy(netid, p, nlen);
	141
	142	/* r_addr: ip/ip6addr with port in dec octets - see RFC 5665 */
35124a09 WAA	143	p = xdr_inline_decode(streamp, 4);
35124a09 WAA	144	if (unlikely(!p))
c9895cb6	145	goto out_free_netid;
35124a09 WAA	146	rlen = be32_to_cpup(p);
	147
	148	p = xdr_inline_decode(streamp, rlen);
	149	if (unlikely(!p))
c9895cb6	150	goto out_free_netid;
35124a09	151
c9895cb6 WAA	152	/* port is ".ABC.DEF", 8 chars max */
c9895cb6 WAA	153	if (rlen > INET6_ADDRSTRLEN + IPV6_SCOPE_ID_LEN + 8) {
ad3d2eed	154	dprintk("%s: Invalid address, length %d\n", __func__,
16b374ca	155	rlen);
c9895cb6	156	goto out_free_netid;
16b374ca	157	}
a75b9df9	158	buf = kmalloc(rlen + 1, gfp_flags);
b9f81057 SF	159	if (!buf) {
b9f81057 SF	160	dprintk("%s: Not enough memory\n", __func__);
c9895cb6	161	goto out_free_netid;
b9f81057	162	}
16b374ca	163	buf[rlen] = '\0';
35124a09	164	memcpy(buf, p, rlen);
16b374ca	165
c9895cb6 WAA	166	/* replace port '.' with '-' */
	167	portstr = strrchr(buf, '.');
	168	if (!portstr) {
	169	dprintk("%s: Failed finding expected dot in port\n",
	170	__func__);
	171	goto out_free_buf;
	172	}
	173	*portstr = '-';
	174
	175	/* find '.' between address and port */
	176	portstr = strrchr(buf, '.');
	177	if (!portstr) {
	178	dprintk("%s: Failed finding expected dot between address and "
	179	"port\n", __func__);
	180	goto out_free_buf;
16b374ca	181	}
c9895cb6	182	*portstr = '\0';
16b374ca	183
14f9a607 WAA	184	da = kzalloc(sizeof(*da), gfp_flags);
14f9a607 WAA	185	if (unlikely(!da))
c9895cb6	186	goto out_free_buf;
14f9a607 WAA	187
	188	INIT_LIST_HEAD(&da->da_node);
	189
17094272	190	if (!rpc_pton(net, buf, portstr-buf, (struct sockaddr *)&da->da_addr,
14f9a607 WAA	191	sizeof(da->da_addr))) {
	192	dprintk("%s: error parsing address %s\n", __func__, buf);
	193	goto out_free_da;
16b374ca AA	194	}
16b374ca AA	195
c9895cb6 WAA	196	portstr++;
c9895cb6 WAA	197	sscanf(portstr, "%d-%d", &tmp[0], &tmp[1]);
16b374ca AA	198	port = htons((tmp[0] << 8) \| (tmp[1]));
16b374ca AA	199
14f9a607	200	switch (da->da_addr.ss_family) {
c9895cb6	201	case AF_INET:
14f9a607 WAA	202	((struct sockaddr_in *)&da->da_addr)->sin_port = port;
14f9a607 WAA	203	da->da_addrlen = sizeof(struct sockaddr_in);
c9895cb6 WAA	204	match_netid = "tcp";
	205	match_netid_len = 3;
	206	break;
	207
	208	case AF_INET6:
14f9a607 WAA	209	((struct sockaddr_in6 *)&da->da_addr)->sin6_port = port;
14f9a607 WAA	210	da->da_addrlen = sizeof(struct sockaddr_in6);
c9895cb6 WAA	211	match_netid = "tcp6";
c9895cb6 WAA	212	match_netid_len = 4;
14f9a607 WAA	213	startsep = "[";
14f9a607 WAA	214	endsep = "]";
c9895cb6 WAA	215	break;
	216
	217	default:
	218	dprintk("%s: unsupported address family: %u\n",
14f9a607 WAA	219	__func__, da->da_addr.ss_family);
14f9a607 WAA	220	goto out_free_da;
c9895cb6 WAA	221	}
	222
	223	if (nlen != match_netid_len \|\| strncmp(netid, match_netid, nlen)) {
	224	dprintk("%s: ERROR: r_netid \"%s\" != \"%s\"\n",
	225	__func__, netid, match_netid);
14f9a607	226	goto out_free_da;
c9895cb6 WAA	227	}
c9895cb6 WAA	228
14f9a607 WAA	229	/* save human readable address */
	230	len = strlen(startsep) + strlen(buf) + strlen(endsep) + 7;
	231	da->da_remotestr = kzalloc(len, gfp_flags);
	232
	233	/* NULL is ok, only used for dprintk */
	234	if (da->da_remotestr)
	235	snprintf(da->da_remotestr, len, "%s%s%s:%u", startsep,
	236	buf, endsep, ntohs(port));
	237
	238	dprintk("%s: Parsed DS addr %s\n", __func__, da->da_remotestr);
	239	kfree(buf);
	240	kfree(netid);
	241	return da;
	242
	243	out_free_da:
	244	kfree(da);
c9895cb6	245	out_free_buf:
14f9a607	246	dprintk("%s: Error parsing DS addr: %s\n", __func__, buf);
16b374ca	247	kfree(buf);
c9895cb6 WAA	248	out_free_netid:
c9895cb6 WAA	249	kfree(netid);
16b374ca	250	out_err:
14f9a607	251	return NULL;
16b374ca AA	252	}
	253
	254	/* Decode opaque device data and return the result */
661373b1 CH	255	struct nfs4_file_layout_dsaddr *
	256	nfs4_fl_alloc_deviceid_node(struct nfs_server server, struct pnfs_device pdev,
	257	gfp_t gfp_flags)
16b374ca	258	{
35124a09	259	int i;
16b374ca AA	260	u32 cnt, num;
16b374ca AA	261	u8 *indexp;
35124a09 WAA	262	__be32 *p;
	263	u8 *stripe_indices;
	264	u8 max_stripe_index;
	265	struct nfs4_file_layout_dsaddr *dsaddr = NULL;
	266	struct xdr_stream stream;
f7da7a12	267	struct xdr_buf buf;
35124a09	268	struct page *scratch;
14f9a607 WAA	269	struct list_head dsaddrs;
14f9a607 WAA	270	struct nfs4_pnfs_ds_addr *da;
35124a09 WAA	271
35124a09 WAA	272	/* set up xdr stream */
a75b9df9	273	scratch = alloc_page(gfp_flags);
35124a09 WAA	274	if (!scratch)
	275	goto out_err;
	276
f7da7a12	277	xdr_init_decode_pages(&stream, &buf, pdev->pages, pdev->pglen);
35124a09	278	xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE);
16b374ca AA	279
16b374ca AA	280	/* Get the stripe count (number of stripe index) */
35124a09 WAA	281	p = xdr_inline_decode(&stream, 4);
	282	if (unlikely(!p))
	283	goto out_err_free_scratch;
	284
	285	cnt = be32_to_cpup(p);
16b374ca AA	286	dprintk("%s stripe count %d\n", __func__, cnt);
16b374ca AA	287	if (cnt > NFS4_PNFS_MAX_STRIPE_CNT) {
a030889a	288	printk(KERN_WARNING "NFS: %s: stripe count %d greater than "
16b374ca AA	289	"supported maximum %d\n", __func__,
16b374ca AA	290	cnt, NFS4_PNFS_MAX_STRIPE_CNT);
35124a09 WAA	291	goto out_err_free_scratch;
	292	}
	293
	294	/* read stripe indices */
a75b9df9	295	stripe_indices = kcalloc(cnt, sizeof(u8), gfp_flags);
35124a09 WAA	296	if (!stripe_indices)
	297	goto out_err_free_scratch;
	298
	299	p = xdr_inline_decode(&stream, cnt << 2);
	300	if (unlikely(!p))
	301	goto out_err_free_stripe_indices;
	302
	303	indexp = &stripe_indices[0];
	304	max_stripe_index = 0;
	305	for (i = 0; i < cnt; i++) {
	306	*indexp = be32_to_cpup(p++);
	307	max_stripe_index = max(max_stripe_index, *indexp);
	308	indexp++;
16b374ca AA	309	}
	310
	311	/* Check the multipath list count */
35124a09 WAA	312	p = xdr_inline_decode(&stream, 4);
	313	if (unlikely(!p))
	314	goto out_err_free_stripe_indices;
	315
	316	num = be32_to_cpup(p);
16b374ca AA	317	dprintk("%s ds_num %u\n", __func__, num);
16b374ca AA	318	if (num > NFS4_PNFS_MAX_MULTI_CNT) {
a030889a	319	printk(KERN_WARNING "NFS: %s: multipath count %d greater than "
16b374ca AA	320	"supported maximum %d\n", __func__,
16b374ca AA	321	num, NFS4_PNFS_MAX_MULTI_CNT);
35124a09	322	goto out_err_free_stripe_indices;
16b374ca	323	}
35124a09 WAA	324
	325	/* validate stripe indices are all < num */
	326	if (max_stripe_index >= num) {
a030889a	327	printk(KERN_WARNING "NFS: %s: stripe index %u >= num ds %u\n",
35124a09 WAA	328	__func__, max_stripe_index, num);
	329	goto out_err_free_stripe_indices;
	330	}
	331
16b374ca AA	332	dsaddr = kzalloc(sizeof(*dsaddr) +
16b374ca AA	333	(sizeof(struct nfs4_pnfs_ds ) (num - 1)),
a75b9df9	334	gfp_flags);
16b374ca	335	if (!dsaddr)
35124a09	336	goto out_err_free_stripe_indices;
16b374ca AA	337
16b374ca AA	338	dsaddr->stripe_count = cnt;
35124a09 WAA	339	dsaddr->stripe_indices = stripe_indices;
35124a09 WAA	340	stripe_indices = NULL;
16b374ca	341	dsaddr->ds_num = num;
661373b1	342	nfs4_init_deviceid_node(&dsaddr->id_node, server, &pdev->dev_id);
16b374ca	343
14f9a607 WAA	344	INIT_LIST_HEAD(&dsaddrs);
14f9a607 WAA	345
16b374ca AA	346	for (i = 0; i < dsaddr->ds_num; i++) {
16b374ca AA	347	int j;
35124a09 WAA	348	u32 mp_count;
	349
	350	p = xdr_inline_decode(&stream, 4);
	351	if (unlikely(!p))
	352	goto out_err_free_deviceid;
16b374ca	353
35124a09	354	mp_count = be32_to_cpup(p); /* multipath count */
35124a09	355	for (j = 0; j < mp_count; j++) {
661373b1	356	da = decode_ds_addr(server->nfs_client->cl_net,
17094272	357	&stream, gfp_flags);
14f9a607 WAA	358	if (da)
	359	list_add_tail(&da->da_node, &dsaddrs);
	360	}
	361	if (list_empty(&dsaddrs)) {
	362	dprintk("%s: no suitable DS addresses found\n",
	363	__func__);
	364	goto out_err_free_deviceid;
	365	}
	366
	367	dsaddr->ds_list[i] = nfs4_pnfs_ds_add(&dsaddrs, gfp_flags);
	368	if (!dsaddr->ds_list[i])
	369	goto out_err_drain_dsaddrs;
	370
	371	/* If DS was already in cache, free ds addrs */
	372	while (!list_empty(&dsaddrs)) {
	373	da = list_first_entry(&dsaddrs,
	374	struct nfs4_pnfs_ds_addr,
	375	da_node);
	376	list_del_init(&da->da_node);
	377	kfree(da->da_remotestr);
	378	kfree(da);
16b374ca AA	379	}
16b374ca AA	380	}
35124a09 WAA	381
35124a09 WAA	382	__free_page(scratch);
16b374ca AA	383	return dsaddr;
16b374ca AA	384
14f9a607 WAA	385	out_err_drain_dsaddrs:
	386	while (!list_empty(&dsaddrs)) {
	387	da = list_first_entry(&dsaddrs, struct nfs4_pnfs_ds_addr,
	388	da_node);
	389	list_del_init(&da->da_node);
	390	kfree(da->da_remotestr);
	391	kfree(da);
	392	}
35124a09	393	out_err_free_deviceid:
16b374ca	394	nfs4_fl_free_deviceid(dsaddr);
35124a09 WAA	395	/* stripe_indicies was part of dsaddr */
	396	goto out_err_free_scratch;
	397	out_err_free_stripe_indices:
	398	kfree(stripe_indices);
	399	out_err_free_scratch:
	400	__free_page(scratch);
16b374ca AA	401	out_err:
	402	dprintk("%s ERROR: returning NULL\n", __func__);
	403	return NULL;
	404	}
	405
ea8eecdd CH	406	void
ea8eecdd CH	407	nfs4_fl_put_deviceid(struct nfs4_file_layout_dsaddr *dsaddr)
16b374ca	408	{
1775bc34	409	nfs4_put_deviceid_node(&dsaddr->id_node);
16b374ca	410	}
cfe7f412 FI	411
	412	/*
	413	* Want res = (offset - layout->pattern_offset)/ layout->stripe_unit
	414	* Then: ((res + fsi) % dsaddr->stripe_count)
	415	*/
	416	u32
	417	nfs4_fl_calc_j_index(struct pnfs_layout_segment *lseg, loff_t offset)
	418	{
	419	struct nfs4_filelayout_segment *flseg = FILELAYOUT_LSEG(lseg);
	420	u64 tmp;
	421
	422	tmp = offset - flseg->pattern_offset;
	423	do_div(tmp, flseg->stripe_unit);
	424	tmp += flseg->first_stripe_index;
	425	return do_div(tmp, flseg->dsaddr->stripe_count);
	426	}
	427
	428	u32
	429	nfs4_fl_calc_ds_index(struct pnfs_layout_segment *lseg, u32 j)
	430	{
	431	return FILELAYOUT_LSEG(lseg)->dsaddr->stripe_indices[j];
	432	}
	433
	434	struct nfs_fh *
	435	nfs4_fl_select_ds_fh(struct pnfs_layout_segment *lseg, u32 j)
	436	{
	437	struct nfs4_filelayout_segment *flseg = FILELAYOUT_LSEG(lseg);
	438	u32 i;
	439
	440	if (flseg->stripe_type == STRIPE_SPARSE) {
	441	if (flseg->num_fh == 1)
	442	i = 0;
	443	else if (flseg->num_fh == 0)
	444	/* Use the MDS OPEN fh set in nfs_read_rpcsetup */
	445	return NULL;
	446	else
	447	i = nfs4_fl_calc_ds_index(lseg, j);
	448	} else
	449	i = j;
	450	return flseg->fh_array[i];
	451	}
	452
c23266d5 AA	453	static void nfs4_wait_ds_connect(struct nfs4_pnfs_ds *ds)
	454	{
	455	might_sleep();
74316201 N	456	wait_on_bit_action(&ds->ds_state, NFS4DS_CONNECTING,
74316201 N	457	nfs_wait_bit_killable, TASK_KILLABLE);
c23266d5 AA	458	}
	459
	460	static void nfs4_clear_ds_conn_bit(struct nfs4_pnfs_ds *ds)
	461	{
4e857c58	462	smp_mb__before_atomic();
c23266d5	463	clear_bit(NFS4DS_CONNECTING, &ds->ds_state);
4e857c58	464	smp_mb__after_atomic();
c23266d5 AA	465	wake_up_bit(&ds->ds_state, NFS4DS_CONNECTING);
	466	}
	467
	468
cfe7f412 FI	469	struct nfs4_pnfs_ds *
	470	nfs4_fl_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx)
	471	{
	472	struct nfs4_file_layout_dsaddr *dsaddr = FILELAYOUT_LSEG(lseg)->dsaddr;
	473	struct nfs4_pnfs_ds *ds = dsaddr->ds_list[ds_idx];
554d458d	474	struct nfs4_deviceid_node *devid = FILELAYOUT_DEVID_NODE(lseg);
52b26a3e	475	struct nfs4_pnfs_ds *ret = ds;
cfe7f412 FI	476
cfe7f412 FI	477	if (ds == NULL) {
a030889a	478	printk(KERN_ERR "NFS: %s: No data server for offset index %d\n",
cfe7f412	479	__func__, ds_idx);
f54bcf2e	480	pnfs_generic_mark_devid_invalid(devid);
52b26a3e	481	goto out;
cfe7f412	482	}
acd65e5b	483	smp_rmb();
c23266d5	484	if (ds->ds_clp)
52b26a3e	485	goto out_test_devid;
cfe7f412	486
c23266d5	487	if (test_and_set_bit(NFS4DS_CONNECTING, &ds->ds_state) == 0) {
568e8c49	488	struct nfs_server *s = NFS_SERVER(lseg->pls_layout->plh_inode);
cfe7f412 FI	489	int err;
cfe7f412 FI	490
568e8c49	491	err = nfs4_ds_connect(s, ds);
52b26a3e	492	if (err)
1dfed273	493	nfs4_mark_deviceid_unavailable(devid);
c23266d5 AA	494	nfs4_clear_ds_conn_bit(ds);
	495	} else {
	496	/* Either ds is connected, or ds is NULL */
	497	nfs4_wait_ds_connect(ds);
cfe7f412	498	}
52b26a3e TM	499	out_test_devid:
	500	if (filelayout_test_devid_unavailable(devid))
	501	ret = NULL;
	502	out:
	503	return ret;
cfe7f412	504	}
98fc685a AA	505
	506	module_param(dataserver_retrans, uint, 0644);
	507	MODULE_PARM_DESC(dataserver_retrans, "The number of times the NFSv4.1 client "
	508	"retries a request before it attempts further "
	509	" recovery action.");
	510	module_param(dataserver_timeo, uint, 0644);
	511	MODULE_PARM_DESC(dataserver_timeo, "The time (in tenths of a second) the "
	512	"NFSv4.1 client waits for a response from a "
	513	" data server before it retries an NFS request.");