[deliverable/linux.git] / fs / nfsd / blocklayout.c

/*
 * Copyright (c) 2014-2016 Christoph Hellwig.
 */
#include <linux/exportfs.h>
#include <linux/genhd.h>
#include <linux/slab.h>
#include <linux/pr.h>

#include <linux/nfsd/debug.h>
#include <scsi/scsi_proto.h>
#include <scsi/scsi_common.h>

#include "blocklayoutxdr.h"
#include "pnfs.h"

#define NFSDDBG_FACILITY	NFSDDBG_PNFS


static __be32
nfsd4_block_proc_layoutget(struct inode *inode, const struct svc_fh *fhp,
		struct nfsd4_layoutget *args)
{
	struct nfsd4_layout_seg *seg = &args->lg_seg;
	struct super_block *sb = inode->i_sb;
	u32 block_size = (1 << inode->i_blkbits);
	struct pnfs_block_extent *bex;
	struct iomap iomap;
	u32 device_generation = 0;
	int error;

	if (seg->offset & (block_size - 1)) {
		dprintk("pnfsd: I/O misaligned\n");
		goto out_layoutunavailable;
	}

	/*
	 * Some clients barf on non-zero block numbers for NONE or INVALID
	 * layouts, so make sure to zero the whole structure.
	 */
	error = -ENOMEM;
	bex = kzalloc(sizeof(*bex), GFP_KERNEL);
	if (!bex)
		goto out_error;
	args->lg_content = bex;

	error = sb->s_export_op->map_blocks(inode, seg->offset, seg->length,
					    &iomap, seg->iomode != IOMODE_READ,
					    &device_generation);
	if (error) {
		if (error == -ENXIO)
			goto out_layoutunavailable;
		goto out_error;
	}

	if (iomap.length < args->lg_minlength) {
		dprintk("pnfsd: extent smaller than minlength\n");
		goto out_layoutunavailable;
	}

	switch (iomap.type) {
	case IOMAP_MAPPED:
		if (seg->iomode == IOMODE_READ)
			bex->es = PNFS_BLOCK_READ_DATA;
		else
			bex->es = PNFS_BLOCK_READWRITE_DATA;
		bex->soff = (iomap.blkno << 9);
		break;
	case IOMAP_UNWRITTEN:
		if (seg->iomode & IOMODE_RW) {
			/*
			 * Crack monkey special case from section 2.3.1.
			 */
			if (args->lg_minlength == 0) {
				dprintk("pnfsd: no soup for you!\n");
				goto out_layoutunavailable;
			}

			bex->es = PNFS_BLOCK_INVALID_DATA;
			bex->soff = (iomap.blkno << 9);
			break;
		}
		/*FALLTHRU*/
	case IOMAP_HOLE:
		if (seg->iomode == IOMODE_READ) {
			bex->es = PNFS_BLOCK_NONE_DATA;
			break;
		}
		/*FALLTHRU*/
	case IOMAP_DELALLOC:
	default:
		WARN(1, "pnfsd: filesystem returned %d extent\n", iomap.type);
		goto out_layoutunavailable;
	}

	error = nfsd4_set_deviceid(&bex->vol_id, fhp, device_generation);
	if (error)
		goto out_error;
	bex->foff = iomap.offset;
	bex->len = iomap.length;

	seg->offset = iomap.offset;
	seg->length = iomap.length;

	dprintk("GET: 0x%llx:0x%llx %d\n", bex->foff, bex->len, bex->es);
	return 0;

out_error:
	seg->length = 0;
	return nfserrno(error);
out_layoutunavailable:
	seg->length = 0;
	return nfserr_layoutunavailable;
}

static __be32
nfsd4_block_commit_blocks(struct inode *inode, struct nfsd4_layoutcommit *lcp,
		struct iomap *iomaps, int nr_iomaps)
{
	loff_t new_size = lcp->lc_last_wr + 1;
	struct iattr iattr = { .ia_valid = 0 };
	int error;

	if (lcp->lc_mtime.tv_nsec == UTIME_NOW ||
	    timespec_compare(&lcp->lc_mtime, &inode->i_mtime) < 0)
		lcp->lc_mtime = current_fs_time(inode->i_sb);
	iattr.ia_valid |= ATTR_ATIME | ATTR_CTIME | ATTR_MTIME;
	iattr.ia_atime = iattr.ia_ctime = iattr.ia_mtime = lcp->lc_mtime;

	if (new_size > i_size_read(inode)) {
		iattr.ia_valid |= ATTR_SIZE;
		iattr.ia_size = new_size;
	}

	error = inode->i_sb->s_export_op->commit_blocks(inode, iomaps,
			nr_iomaps, &iattr);
	kfree(iomaps);
	return nfserrno(error);
}

#ifdef CONFIG_NFSD_BLOCKLAYOUT
static int
nfsd4_block_get_device_info_simple(struct super_block *sb,
		struct nfsd4_getdeviceinfo *gdp)
{
	struct pnfs_block_deviceaddr *dev;
	struct pnfs_block_volume *b;

	dev = kzalloc(sizeof(struct pnfs_block_deviceaddr) +
		      sizeof(struct pnfs_block_volume), GFP_KERNEL);
	if (!dev)
		return -ENOMEM;
	gdp->gd_device = dev;

	dev->nr_volumes = 1;
	b = &dev->volumes[0];

	b->type = PNFS_BLOCK_VOLUME_SIMPLE;
	b->simple.sig_len = PNFS_BLOCK_UUID_LEN;
	return sb->s_export_op->get_uuid(sb, b->simple.sig, &b->simple.sig_len,
			&b->simple.offset);
}

static __be32
nfsd4_block_proc_getdeviceinfo(struct super_block *sb,
		struct svc_rqst *rqstp,
		struct nfs4_client *clp,
		struct nfsd4_getdeviceinfo *gdp)
{
	if (sb->s_bdev != sb->s_bdev->bd_contains)
		return nfserr_inval;
	return nfserrno(nfsd4_block_get_device_info_simple(sb, gdp));
}

static __be32
nfsd4_block_proc_layoutcommit(struct inode *inode,
		struct nfsd4_layoutcommit *lcp)
{
	struct iomap *iomaps;
	int nr_iomaps;

	nr_iomaps = nfsd4_block_decode_layoutupdate(lcp->lc_up_layout,
			lcp->lc_up_len, &iomaps, 1 << inode->i_blkbits);
	if (nr_iomaps < 0)
		return nfserrno(nr_iomaps);

	return nfsd4_block_commit_blocks(inode, lcp, iomaps, nr_iomaps);
}

const struct nfsd4_layout_ops bl_layout_ops = {
	/*
	 * Pretend that we send notification to the client.  This is a blatant
	 * lie to force recent Linux clients to cache our device IDs.
	 * We rarely ever change the device ID, so the harm of leaking deviceids
	 * for a while isn't too bad.  Unfortunately RFC5661 is a complete mess
	 * in this regard, but I filed errata 4119 for this a while ago, and
	 * hopefully the Linux client will eventually start caching deviceids
	 * without this again.
	 */
	.notify_types		=
			NOTIFY_DEVICEID4_DELETE | NOTIFY_DEVICEID4_CHANGE,
	.proc_getdeviceinfo	= nfsd4_block_proc_getdeviceinfo,
	.encode_getdeviceinfo	= nfsd4_block_encode_getdeviceinfo,
	.proc_layoutget		= nfsd4_block_proc_layoutget,
	.encode_layoutget	= nfsd4_block_encode_layoutget,
	.proc_layoutcommit	= nfsd4_block_proc_layoutcommit,
};
#endif /* CONFIG_NFSD_BLOCKLAYOUT */

#ifdef CONFIG_NFSD_SCSILAYOUT
static int nfsd4_scsi_identify_device(struct block_device *bdev,
		struct pnfs_block_volume *b)
{
	struct request_queue *q = bdev->bd_disk->queue;
	struct request *rq;
	size_t bufflen = 252, len, id_len;
	u8 *buf, *d, type, assoc;
	int error;

	buf = kzalloc(bufflen, GFP_KERNEL);
	if (!buf)
		return -ENOMEM;

	rq = blk_get_request(q, READ, GFP_KERNEL);
	if (IS_ERR(rq)) {
		error = -ENOMEM;
		goto out_free_buf;
	}
	blk_rq_set_block_pc(rq);

	error = blk_rq_map_kern(q, rq, buf, bufflen, GFP_KERNEL);
	if (error)
		goto out_put_request;

	rq->cmd[0] = INQUIRY;
	rq->cmd[1] = 1;
	rq->cmd[2] = 0x83;
	rq->cmd[3] = bufflen >> 8;
	rq->cmd[4] = bufflen & 0xff;
	rq->cmd_len = COMMAND_SIZE(INQUIRY);

	error = blk_execute_rq(rq->q, NULL, rq, 1);
	if (error) {
		pr_err("pNFS: INQUIRY 0x83 failed with: %x\n",
			rq->errors);
		goto out_put_request;
	}

	len = (buf[2] << 8) + buf[3] + 4;
	if (len > bufflen) {
		pr_err("pNFS: INQUIRY 0x83 response invalid (len = %zd)\n",
			len);
		goto out_put_request;
	}

	d = buf + 4;
	for (d = buf + 4; d < buf + len; d += id_len + 4) {
		id_len = d[3];
		type = d[1] & 0xf;
		assoc = (d[1] >> 4) & 0x3;

		/*
		 * We only care about a EUI-64 and NAA designator types
		 * with LU association.
		 */
		if (assoc != 0x00)
			continue;
		if (type != 0x02 && type != 0x03)
			continue;
		if (id_len != 8 && id_len != 12 && id_len != 16)
			continue;

		b->scsi.code_set = PS_CODE_SET_BINARY;
		b->scsi.designator_type = type == 0x02 ?
			PS_DESIGNATOR_EUI64 : PS_DESIGNATOR_NAA;
		b->scsi.designator_len = id_len;
		memcpy(b->scsi.designator, d + 4, id_len);

		/*
		 * If we found a 8 or 12 byte descriptor continue on to
		 * see if a 16 byte one is available.  If we find a
		 * 16 byte descriptor we're done.
		 */
		if (id_len == 16)
			break;
	}

out_put_request:
	blk_put_request(rq);
out_free_buf:
	kfree(buf);
	return error;
}

#define NFSD_MDS_PR_KEY		0x0100000000000000ULL

/*
 * We use the client ID as a unique key for the reservations.
 * This allows us to easily fence a client when recalls fail.
 */
static u64 nfsd4_scsi_pr_key(struct nfs4_client *clp)
{
	return ((u64)clp->cl_clientid.cl_boot << 32) | clp->cl_clientid.cl_id;
}

static int
nfsd4_block_get_device_info_scsi(struct super_block *sb,
		struct nfs4_client *clp,
		struct nfsd4_getdeviceinfo *gdp)
{
	struct pnfs_block_deviceaddr *dev;
	struct pnfs_block_volume *b;
	const struct pr_ops *ops;
	int error;

	dev = kzalloc(sizeof(struct pnfs_block_deviceaddr) +
		      sizeof(struct pnfs_block_volume), GFP_KERNEL);
	if (!dev)
		return -ENOMEM;
	gdp->gd_device = dev;

	dev->nr_volumes = 1;
	b = &dev->volumes[0];

	b->type = PNFS_BLOCK_VOLUME_SCSI;
	b->scsi.pr_key = nfsd4_scsi_pr_key(clp);

	error = nfsd4_scsi_identify_device(sb->s_bdev, b);
	if (error)
		return error;

	ops = sb->s_bdev->bd_disk->fops->pr_ops;
	if (!ops) {
		pr_err("pNFS: device %s does not support PRs.\n",
			sb->s_id);
		return -EINVAL;
	}

	error = ops->pr_register(sb->s_bdev, 0, NFSD_MDS_PR_KEY, true);
	if (error) {
		pr_err("pNFS: failed to register key for device %s.\n",
			sb->s_id);
		return -EINVAL;
	}

	error = ops->pr_reserve(sb->s_bdev, NFSD_MDS_PR_KEY,
			PR_EXCLUSIVE_ACCESS_REG_ONLY, 0);
	if (error) {
		pr_err("pNFS: failed to reserve device %s.\n",
			sb->s_id);
		return -EINVAL;
	}

	return 0;
}

static __be32
nfsd4_scsi_proc_getdeviceinfo(struct super_block *sb,
		struct svc_rqst *rqstp,
		struct nfs4_client *clp,
		struct nfsd4_getdeviceinfo *gdp)
{
	if (sb->s_bdev != sb->s_bdev->bd_contains)
		return nfserr_inval;
	return nfserrno(nfsd4_block_get_device_info_scsi(sb, clp, gdp));
}
static __be32
nfsd4_scsi_proc_layoutcommit(struct inode *inode,
		struct nfsd4_layoutcommit *lcp)
{
	struct iomap *iomaps;
	int nr_iomaps;

	nr_iomaps = nfsd4_scsi_decode_layoutupdate(lcp->lc_up_layout,
			lcp->lc_up_len, &iomaps, 1 << inode->i_blkbits);
	if (nr_iomaps < 0)
		return nfserrno(nr_iomaps);

	return nfsd4_block_commit_blocks(inode, lcp, iomaps, nr_iomaps);
}

static void
nfsd4_scsi_fence_client(struct nfs4_layout_stateid *ls)
{
	struct nfs4_client *clp = ls->ls_stid.sc_client;
	struct block_device *bdev = ls->ls_file->f_path.mnt->mnt_sb->s_bdev;

	bdev->bd_disk->fops->pr_ops->pr_preempt(bdev, NFSD_MDS_PR_KEY,
			nfsd4_scsi_pr_key(clp), 0, true);
}

const struct nfsd4_layout_ops scsi_layout_ops = {
	/*
	 * Pretend that we send notification to the client.  This is a blatant
	 * lie to force recent Linux clients to cache our device IDs.
	 * We rarely ever change the device ID, so the harm of leaking deviceids
	 * for a while isn't too bad.  Unfortunately RFC5661 is a complete mess
	 * in this regard, but I filed errata 4119 for this a while ago, and
	 * hopefully the Linux client will eventually start caching deviceids
	 * without this again.
	 */
	.notify_types		=
			NOTIFY_DEVICEID4_DELETE | NOTIFY_DEVICEID4_CHANGE,
	.proc_getdeviceinfo	= nfsd4_scsi_proc_getdeviceinfo,
	.encode_getdeviceinfo	= nfsd4_block_encode_getdeviceinfo,
	.proc_layoutget		= nfsd4_block_proc_layoutget,
	.encode_layoutget	= nfsd4_block_encode_layoutget,
	.proc_layoutcommit	= nfsd4_scsi_proc_layoutcommit,
	.fence_client		= nfsd4_scsi_fence_client,
};
#endif /* CONFIG_NFSD_SCSILAYOUT */
Commit	Line	Data
8650b8a0	1	/*
f99d4fbd	2	* Copyright (c) 2014-2016 Christoph Hellwig.
8650b8a0 CH	3	*/
	4	#include <linux/exportfs.h>
	5	#include <linux/genhd.h>
	6	#include <linux/slab.h>
f99d4fbd	7	#include <linux/pr.h>
8650b8a0 CH	8
8650b8a0 CH	9	#include <linux/nfsd/debug.h>
f99d4fbd CH	10	#include <scsi/scsi_proto.h>
f99d4fbd CH	11	#include <scsi/scsi_common.h>
8650b8a0 CH	12
	13	#include "blocklayoutxdr.h"
	14	#include "pnfs.h"
	15
	16	#define NFSDDBG_FACILITY NFSDDBG_PNFS
	17
	18
8650b8a0 CH	19	static __be32
	20	nfsd4_block_proc_layoutget(struct inode inode, const struct svc_fh fhp,
	21	struct nfsd4_layoutget *args)
	22	{
	23	struct nfsd4_layout_seg *seg = &args->lg_seg;
	24	struct super_block *sb = inode->i_sb;
	25	u32 block_size = (1 << inode->i_blkbits);
	26	struct pnfs_block_extent *bex;
	27	struct iomap iomap;
	28	u32 device_generation = 0;
	29	int error;
	30
8650b8a0 CH	31	if (seg->offset & (block_size - 1)) {
	32	dprintk("pnfsd: I/O misaligned\n");
	33	goto out_layoutunavailable;
	34	}
	35
	36	/*
	37	* Some clients barf on non-zero block numbers for NONE or INVALID
	38	* layouts, so make sure to zero the whole structure.
	39	*/
	40	error = -ENOMEM;
	41	bex = kzalloc(sizeof(*bex), GFP_KERNEL);
	42	if (!bex)
	43	goto out_error;
	44	args->lg_content = bex;
	45
	46	error = sb->s_export_op->map_blocks(inode, seg->offset, seg->length,
	47	&iomap, seg->iomode != IOMODE_READ,
	48	&device_generation);
	49	if (error) {
	50	if (error == -ENXIO)
	51	goto out_layoutunavailable;
	52	goto out_error;
	53	}
	54
	55	if (iomap.length < args->lg_minlength) {
	56	dprintk("pnfsd: extent smaller than minlength\n");
	57	goto out_layoutunavailable;
	58	}
	59
	60	switch (iomap.type) {
	61	case IOMAP_MAPPED:
	62	if (seg->iomode == IOMODE_READ)
	63	bex->es = PNFS_BLOCK_READ_DATA;
	64	else
	65	bex->es = PNFS_BLOCK_READWRITE_DATA;
	66	bex->soff = (iomap.blkno << 9);
	67	break;
	68	case IOMAP_UNWRITTEN:
	69	if (seg->iomode & IOMODE_RW) {
	70	/*
	71	* Crack monkey special case from section 2.3.1.
	72	*/
	73	if (args->lg_minlength == 0) {
	74	dprintk("pnfsd: no soup for you!\n");
	75	goto out_layoutunavailable;
	76	}
	77
	78	bex->es = PNFS_BLOCK_INVALID_DATA;
	79	bex->soff = (iomap.blkno << 9);
	80	break;
	81	}
	82	/FALLTHRU/
	83	case IOMAP_HOLE:
	84	if (seg->iomode == IOMODE_READ) {
	85	bex->es = PNFS_BLOCK_NONE_DATA;
	86	break;
	87	}
	88	/FALLTHRU/
	89	case IOMAP_DELALLOC:
	90	default:
	91	WARN(1, "pnfsd: filesystem returned %d extent\n", iomap.type);
	92	goto out_layoutunavailable;
	93	}
	94
95	error = nfsd4_set_deviceid(&bex->vol_id, fhp, device_generation);
96	if (error)
97	goto out_error;
98	bex->foff = iomap.offset;
99	bex->len = iomap.length;
100
101	seg->offset = iomap.offset;
102	seg->length = iomap.length;
103
85369523	104	dprintk("GET: 0x%llx:0x%llx %d\n", bex->foff, bex->len, bex->es);
8650b8a0 CH	105	return 0;
	106
	107	out_error:
	108	seg->length = 0;
	109	return nfserrno(error);
	110	out_layoutunavailable:
	111	seg->length = 0;
	112	return nfserr_layoutunavailable;
	113	}
	114
	115	static __be32
368248ee CH	116	nfsd4_block_commit_blocks(struct inode inode, struct nfsd4_layoutcommit lcp,
368248ee CH	117	struct iomap *iomaps, int nr_iomaps)
8650b8a0 CH	118	{
	119	loff_t new_size = lcp->lc_last_wr + 1;
	120	struct iattr iattr = { .ia_valid = 0 };
8650b8a0 CH	121	int error;
8650b8a0 CH	122
8650b8a0 CH	123	if (lcp->lc_mtime.tv_nsec == UTIME_NOW \|\|
	124	timespec_compare(&lcp->lc_mtime, &inode->i_mtime) < 0)
	125	lcp->lc_mtime = current_fs_time(inode->i_sb);
	126	iattr.ia_valid \|= ATTR_ATIME \| ATTR_CTIME \| ATTR_MTIME;
	127	iattr.ia_atime = iattr.ia_ctime = iattr.ia_mtime = lcp->lc_mtime;
	128
	129	if (new_size > i_size_read(inode)) {
	130	iattr.ia_valid \|= ATTR_SIZE;
	131	iattr.ia_size = new_size;
	132	}
	133
	134	error = inode->i_sb->s_export_op->commit_blocks(inode, iomaps,
	135	nr_iomaps, &iattr);
	136	kfree(iomaps);
	137	return nfserrno(error);
	138	}
	139
368248ee CH	140	#ifdef CONFIG_NFSD_BLOCKLAYOUT
	141	static int
	142	nfsd4_block_get_device_info_simple(struct super_block *sb,
	143	struct nfsd4_getdeviceinfo *gdp)
	144	{
	145	struct pnfs_block_deviceaddr *dev;
	146	struct pnfs_block_volume *b;
	147
	148	dev = kzalloc(sizeof(struct pnfs_block_deviceaddr) +
	149	sizeof(struct pnfs_block_volume), GFP_KERNEL);
	150	if (!dev)
	151	return -ENOMEM;
	152	gdp->gd_device = dev;
	153
	154	dev->nr_volumes = 1;
	155	b = &dev->volumes[0];
	156
	157	b->type = PNFS_BLOCK_VOLUME_SIMPLE;
	158	b->simple.sig_len = PNFS_BLOCK_UUID_LEN;
	159	return sb->s_export_op->get_uuid(sb, b->simple.sig, &b->simple.sig_len,
	160	&b->simple.offset);
	161	}
	162
	163	static __be32
	164	nfsd4_block_proc_getdeviceinfo(struct super_block *sb,
d7c920d1	165	struct svc_rqst *rqstp,
f99d4fbd	166	struct nfs4_client *clp,
368248ee CH	167	struct nfsd4_getdeviceinfo *gdp)
	168	{
	169	if (sb->s_bdev != sb->s_bdev->bd_contains)
	170	return nfserr_inval;
	171	return nfserrno(nfsd4_block_get_device_info_simple(sb, gdp));
	172	}
	173
	174	static __be32
	175	nfsd4_block_proc_layoutcommit(struct inode *inode,
	176	struct nfsd4_layoutcommit *lcp)
	177	{
	178	struct iomap *iomaps;
	179	int nr_iomaps;
	180
	181	nr_iomaps = nfsd4_block_decode_layoutupdate(lcp->lc_up_layout,
	182	lcp->lc_up_len, &iomaps, 1 << inode->i_blkbits);
	183	if (nr_iomaps < 0)
	184	return nfserrno(nr_iomaps);
	185
	186	return nfsd4_block_commit_blocks(inode, lcp, iomaps, nr_iomaps);
	187	}
	188
8650b8a0	189	const struct nfsd4_layout_ops bl_layout_ops = {
40cdc7a5 CH	190	/*
	191	* Pretend that we send notification to the client. This is a blatant
	192	* lie to force recent Linux clients to cache our device IDs.
	193	* We rarely ever change the device ID, so the harm of leaking deviceids
	194	* for a while isn't too bad. Unfortunately RFC5661 is a complete mess
	195	* in this regard, but I filed errata 4119 for this a while ago, and
	196	* hopefully the Linux client will eventually start caching deviceids
	197	* without this again.
	198	*/
	199	.notify_types =
	200	NOTIFY_DEVICEID4_DELETE \| NOTIFY_DEVICEID4_CHANGE,
8650b8a0 CH	201	.proc_getdeviceinfo = nfsd4_block_proc_getdeviceinfo,
	202	.encode_getdeviceinfo = nfsd4_block_encode_getdeviceinfo,
	203	.proc_layoutget = nfsd4_block_proc_layoutget,
	204	.encode_layoutget = nfsd4_block_encode_layoutget,
	205	.proc_layoutcommit = nfsd4_block_proc_layoutcommit,
	206	};
368248ee	207	#endif /* CONFIG_NFSD_BLOCKLAYOUT */
f99d4fbd CH	208
	209	#ifdef CONFIG_NFSD_SCSILAYOUT
	210	static int nfsd4_scsi_identify_device(struct block_device *bdev,
	211	struct pnfs_block_volume *b)
	212	{
	213	struct request_queue *q = bdev->bd_disk->queue;
	214	struct request *rq;
	215	size_t bufflen = 252, len, id_len;
	216	u8 buf, d, type, assoc;
	217	int error;
	218
	219	buf = kzalloc(bufflen, GFP_KERNEL);
	220	if (!buf)
	221	return -ENOMEM;
	222
	223	rq = blk_get_request(q, READ, GFP_KERNEL);
	224	if (IS_ERR(rq)) {
	225	error = -ENOMEM;
	226	goto out_free_buf;
	227	}
	228	blk_rq_set_block_pc(rq);
	229
	230	error = blk_rq_map_kern(q, rq, buf, bufflen, GFP_KERNEL);
	231	if (error)
	232	goto out_put_request;
	233
	234	rq->cmd[0] = INQUIRY;
	235	rq->cmd[1] = 1;
	236	rq->cmd[2] = 0x83;
	237	rq->cmd[3] = bufflen >> 8;
	238	rq->cmd[4] = bufflen & 0xff;
	239	rq->cmd_len = COMMAND_SIZE(INQUIRY);
	240
	241	error = blk_execute_rq(rq->q, NULL, rq, 1);
	242	if (error) {
	243	pr_err("pNFS: INQUIRY 0x83 failed with: %x\n",
	244	rq->errors);
	245	goto out_put_request;
	246	}
	247
	248	len = (buf[2] << 8) + buf[3] + 4;
	249	if (len > bufflen) {
	250	pr_err("pNFS: INQUIRY 0x83 response invalid (len = %zd)\n",
	251	len);
	252	goto out_put_request;
	253	}
	254
	255	d = buf + 4;
	256	for (d = buf + 4; d < buf + len; d += id_len + 4) {
	257	id_len = d[3];
	258	type = d[1] & 0xf;
	259	assoc = (d[1] >> 4) & 0x3;
	260
	261	/*
	262	* We only care about a EUI-64 and NAA designator types
	263	* with LU association.
	264	*/
	265	if (assoc != 0x00)
	266	continue;
	267	if (type != 0x02 && type != 0x03)
	268	continue;
	269	if (id_len != 8 && id_len != 12 && id_len != 16)
	270	continue;
	271
272	b->scsi.code_set = PS_CODE_SET_BINARY;
273	b->scsi.designator_type = type == 0x02 ?
274	PS_DESIGNATOR_EUI64 : PS_DESIGNATOR_NAA;
275	b->scsi.designator_len = id_len;
276	memcpy(b->scsi.designator, d + 4, id_len);
277
278	/*
279	* If we found a 8 or 12 byte descriptor continue on to
280	* see if a 16 byte one is available. If we find a
281	* 16 byte descriptor we're done.
282	*/
283	if (id_len == 16)
284	break;
285	}
286
287	out_put_request:
288	blk_put_request(rq);
289	out_free_buf:
290	kfree(buf);
291	return error;
292	}
293
eee93016	294	#define NFSD_MDS_PR_KEY 0x0100000000000000ULL
f99d4fbd CH	295
	296	/*
	297	* We use the client ID as a unique key for the reservations.
	298	* This allows us to easily fence a client when recalls fail.
	299	*/
	300	static u64 nfsd4_scsi_pr_key(struct nfs4_client *clp)
	301	{
	302	return ((u64)clp->cl_clientid.cl_boot << 32) \| clp->cl_clientid.cl_id;
	303	}
	304
	305	static int
	306	nfsd4_block_get_device_info_scsi(struct super_block *sb,
	307	struct nfs4_client *clp,
	308	struct nfsd4_getdeviceinfo *gdp)
	309	{
	310	struct pnfs_block_deviceaddr *dev;
	311	struct pnfs_block_volume *b;
	312	const struct pr_ops *ops;
	313	int error;
	314
	315	dev = kzalloc(sizeof(struct pnfs_block_deviceaddr) +
	316	sizeof(struct pnfs_block_volume), GFP_KERNEL);
	317	if (!dev)
	318	return -ENOMEM;
	319	gdp->gd_device = dev;
	320
	321	dev->nr_volumes = 1;
	322	b = &dev->volumes[0];
	323
	324	b->type = PNFS_BLOCK_VOLUME_SCSI;
	325	b->scsi.pr_key = nfsd4_scsi_pr_key(clp);
	326
	327	error = nfsd4_scsi_identify_device(sb->s_bdev, b);
	328	if (error)
	329	return error;
	330
	331	ops = sb->s_bdev->bd_disk->fops->pr_ops;
	332	if (!ops) {
	333	pr_err("pNFS: device %s does not support PRs.\n",
	334	sb->s_id);
	335	return -EINVAL;
	336	}
	337
	338	error = ops->pr_register(sb->s_bdev, 0, NFSD_MDS_PR_KEY, true);
	339	if (error) {
	340	pr_err("pNFS: failed to register key for device %s.\n",
	341	sb->s_id);
	342	return -EINVAL;
	343	}
	344
	345	error = ops->pr_reserve(sb->s_bdev, NFSD_MDS_PR_KEY,
	346	PR_EXCLUSIVE_ACCESS_REG_ONLY, 0);
	347	if (error) {
	348	pr_err("pNFS: failed to reserve device %s.\n",
	349	sb->s_id);
	350	return -EINVAL;
	351	}
	352
	353	return 0;
	354	}
	355
	356	static __be32
	357	nfsd4_scsi_proc_getdeviceinfo(struct super_block *sb,
d7c920d1	358	struct svc_rqst *rqstp,
f99d4fbd CH	359	struct nfs4_client *clp,
	360	struct nfsd4_getdeviceinfo *gdp)
	361	{
	362	if (sb->s_bdev != sb->s_bdev->bd_contains)
	363	return nfserr_inval;
	364	return nfserrno(nfsd4_block_get_device_info_scsi(sb, clp, gdp));
	365	}
	366	static __be32
	367	nfsd4_scsi_proc_layoutcommit(struct inode *inode,
	368	struct nfsd4_layoutcommit *lcp)
	369	{
	370	struct iomap *iomaps;
	371	int nr_iomaps;
	372
	373	nr_iomaps = nfsd4_scsi_decode_layoutupdate(lcp->lc_up_layout,
	374	lcp->lc_up_len, &iomaps, 1 << inode->i_blkbits);
	375	if (nr_iomaps < 0)
	376	return nfserrno(nr_iomaps);
	377
	378	return nfsd4_block_commit_blocks(inode, lcp, iomaps, nr_iomaps);
	379	}
	380
	381	static void
	382	nfsd4_scsi_fence_client(struct nfs4_layout_stateid *ls)
	383	{
	384	struct nfs4_client *clp = ls->ls_stid.sc_client;
	385	struct block_device *bdev = ls->ls_file->f_path.mnt->mnt_sb->s_bdev;
	386
	387	bdev->bd_disk->fops->pr_ops->pr_preempt(bdev, NFSD_MDS_PR_KEY,
	388	nfsd4_scsi_pr_key(clp), 0, true);
	389	}
	390
	391	const struct nfsd4_layout_ops scsi_layout_ops = {
	392	/*
	393	* Pretend that we send notification to the client. This is a blatant
	394	* lie to force recent Linux clients to cache our device IDs.
	395	* We rarely ever change the device ID, so the harm of leaking deviceids
	396	* for a while isn't too bad. Unfortunately RFC5661 is a complete mess
	397	* in this regard, but I filed errata 4119 for this a while ago, and
	398	* hopefully the Linux client will eventually start caching deviceids
	399	* without this again.
	400	*/
	401	.notify_types =
	402	NOTIFY_DEVICEID4_DELETE \| NOTIFY_DEVICEID4_CHANGE,
	403	.proc_getdeviceinfo = nfsd4_scsi_proc_getdeviceinfo,
	404	.encode_getdeviceinfo = nfsd4_block_encode_getdeviceinfo,
	405	.proc_layoutget = nfsd4_block_proc_layoutget,
	406	.encode_layoutget = nfsd4_block_encode_layoutget,
	407	.proc_layoutcommit = nfsd4_scsi_proc_layoutcommit,
	408	.fence_client = nfsd4_scsi_fence_client,
	409	};
	410	#endif /* CONFIG_NFSD_SCSILAYOUT */