[linux-2.6-block.git] / fs / nfsd / blocklayout.c

// SPDX-License-Identifier: GPL-2.0
/*
 * Copyright (c) 2014-2016 Christoph Hellwig.
 */
#include <linux/exportfs.h>
#include <linux/iomap.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 <scsi/scsi_request.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 = i_blocksize(inode);
	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.addr;
		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.addr;
			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 ||
	    timespec64_compare(&lcp->lc_mtime, &inode->i_mtime) < 0)
		lcp->lc_mtime = current_time(inode);
	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, i_blocksize(inode));
	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;
	struct scsi_request *req;
	/*
	 * The allocation length (passed in bytes 3 and 4 of the INQUIRY
	 * command descriptor block) specifies the number of bytes that have
	 * been allocated for the data-in buffer.
	 * 252 is the highest one-byte value that is a multiple of 4.
	 * 65532 is the highest two-byte value that is a multiple of 4.
	 */
	size_t bufflen = 252, maxlen = 65532, len, id_len;
	u8 *buf, *d, type, assoc;
	int retries = 1, error;

	if (WARN_ON_ONCE(!blk_queue_scsi_passthrough(q)))
		return -EINVAL;

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

	rq = blk_get_request(q, REQ_OP_SCSI_IN, 0);
	if (IS_ERR(rq)) {
		error = -ENOMEM;
		goto out_free_buf;
	}
	req = scsi_req(rq);

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

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

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

	len = (buf[2] << 8) + buf[3] + 4;
	if (len > bufflen) {
		if (len <= maxlen && retries--) {
			blk_put_request(rq);
			kfree(buf);
			bufflen = len;
			goto again;
		}
		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, i_blocksize(inode));
	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
b2441318	1	// SPDX-License-Identifier: GPL-2.0
8650b8a0	2	/*
f99d4fbd	3	* Copyright (c) 2014-2016 Christoph Hellwig.
8650b8a0 CH	4	*/
8650b8a0 CH	5	#include <linux/exportfs.h>
199a31c6	6	#include <linux/iomap.h>
8650b8a0 CH	7	#include <linux/genhd.h>
8650b8a0 CH	8	#include <linux/slab.h>
f99d4fbd	9	#include <linux/pr.h>
8650b8a0 CH	10
8650b8a0 CH	11	#include <linux/nfsd/debug.h>
f99d4fbd CH	12	#include <scsi/scsi_proto.h>
f99d4fbd CH	13	#include <scsi/scsi_common.h>
82ed4db4	14	#include <scsi/scsi_request.h>
8650b8a0 CH	15
	16	#include "blocklayoutxdr.h"
	17	#include "pnfs.h"
	18
	19	#define NFSDDBG_FACILITY NFSDDBG_PNFS
	20
	21
8650b8a0 CH	22	static __be32
	23	nfsd4_block_proc_layoutget(struct inode inode, const struct svc_fh fhp,
	24	struct nfsd4_layoutget *args)
	25	{
	26	struct nfsd4_layout_seg *seg = &args->lg_seg;
	27	struct super_block *sb = inode->i_sb;
93407472	28	u32 block_size = i_blocksize(inode);
8650b8a0 CH	29	struct pnfs_block_extent *bex;
	30	struct iomap iomap;
	31	u32 device_generation = 0;
	32	int error;
	33
8650b8a0 CH	34	if (seg->offset & (block_size - 1)) {
	35	dprintk("pnfsd: I/O misaligned\n");
	36	goto out_layoutunavailable;
	37	}
	38
	39	/*
	40	* Some clients barf on non-zero block numbers for NONE or INVALID
	41	* layouts, so make sure to zero the whole structure.
	42	*/
	43	error = -ENOMEM;
	44	bex = kzalloc(sizeof(*bex), GFP_KERNEL);
	45	if (!bex)
	46	goto out_error;
	47	args->lg_content = bex;
	48
	49	error = sb->s_export_op->map_blocks(inode, seg->offset, seg->length,
	50	&iomap, seg->iomode != IOMODE_READ,
	51	&device_generation);
	52	if (error) {
	53	if (error == -ENXIO)
	54	goto out_layoutunavailable;
	55	goto out_error;
	56	}
	57
	58	if (iomap.length < args->lg_minlength) {
	59	dprintk("pnfsd: extent smaller than minlength\n");
	60	goto out_layoutunavailable;
	61	}
	62
	63	switch (iomap.type) {
	64	case IOMAP_MAPPED:
	65	if (seg->iomode == IOMODE_READ)
	66	bex->es = PNFS_BLOCK_READ_DATA;
	67	else
	68	bex->es = PNFS_BLOCK_READWRITE_DATA;
19fe5f64	69	bex->soff = iomap.addr;
8650b8a0 CH	70	break;
	71	case IOMAP_UNWRITTEN:
	72	if (seg->iomode & IOMODE_RW) {
	73	/*
	74	* Crack monkey special case from section 2.3.1.
	75	*/
	76	if (args->lg_minlength == 0) {
	77	dprintk("pnfsd: no soup for you!\n");
	78	goto out_layoutunavailable;
	79	}
	80
	81	bex->es = PNFS_BLOCK_INVALID_DATA;
19fe5f64	82	bex->soff = iomap.addr;
8650b8a0 CH	83	break;
	84	}
	85	/FALLTHRU/
	86	case IOMAP_HOLE:
	87	if (seg->iomode == IOMODE_READ) {
	88	bex->es = PNFS_BLOCK_NONE_DATA;
	89	break;
	90	}
	91	/FALLTHRU/
	92	case IOMAP_DELALLOC:
	93	default:
	94	WARN(1, "pnfsd: filesystem returned %d extent\n", iomap.type);
	95	goto out_layoutunavailable;
	96	}
	97
	98	error = nfsd4_set_deviceid(&bex->vol_id, fhp, device_generation);
	99	if (error)
	100	goto out_error;
	101	bex->foff = iomap.offset;
	102	bex->len = iomap.length;
	103
	104	seg->offset = iomap.offset;
	105	seg->length = iomap.length;
	106
85369523	107	dprintk("GET: 0x%llx:0x%llx %d\n", bex->foff, bex->len, bex->es);
8650b8a0 CH	108	return 0;
	109
	110	out_error:
	111	seg->length = 0;
	112	return nfserrno(error);
	113	out_layoutunavailable:
	114	seg->length = 0;
	115	return nfserr_layoutunavailable;
	116	}
	117
	118	static __be32
368248ee CH	119	nfsd4_block_commit_blocks(struct inode inode, struct nfsd4_layoutcommit lcp,
368248ee CH	120	struct iomap *iomaps, int nr_iomaps)
8650b8a0 CH	121	{
	122	loff_t new_size = lcp->lc_last_wr + 1;
	123	struct iattr iattr = { .ia_valid = 0 };
8650b8a0 CH	124	int error;
8650b8a0 CH	125
8650b8a0	126	if (lcp->lc_mtime.tv_nsec == UTIME_NOW \|\|
bdba5368 BF	127	timespec64_compare(&lcp->lc_mtime, &inode->i_mtime) < 0)
bdba5368 BF	128	lcp->lc_mtime = current_time(inode);
8650b8a0	129	iattr.ia_valid \|= ATTR_ATIME \| ATTR_CTIME \| ATTR_MTIME;
bdba5368	130	iattr.ia_atime = iattr.ia_ctime = iattr.ia_mtime = lcp->lc_mtime;
8650b8a0 CH	131
	132	if (new_size > i_size_read(inode)) {
	133	iattr.ia_valid \|= ATTR_SIZE;
	134	iattr.ia_size = new_size;
	135	}
	136
	137	error = inode->i_sb->s_export_op->commit_blocks(inode, iomaps,
	138	nr_iomaps, &iattr);
	139	kfree(iomaps);
	140	return nfserrno(error);
	141	}
	142
368248ee CH	143	#ifdef CONFIG_NFSD_BLOCKLAYOUT
	144	static int
	145	nfsd4_block_get_device_info_simple(struct super_block *sb,
	146	struct nfsd4_getdeviceinfo *gdp)
	147	{
	148	struct pnfs_block_deviceaddr *dev;
	149	struct pnfs_block_volume *b;
	150
	151	dev = kzalloc(sizeof(struct pnfs_block_deviceaddr) +
	152	sizeof(struct pnfs_block_volume), GFP_KERNEL);
	153	if (!dev)
	154	return -ENOMEM;
	155	gdp->gd_device = dev;
	156
	157	dev->nr_volumes = 1;
	158	b = &dev->volumes[0];
	159
	160	b->type = PNFS_BLOCK_VOLUME_SIMPLE;
	161	b->simple.sig_len = PNFS_BLOCK_UUID_LEN;
	162	return sb->s_export_op->get_uuid(sb, b->simple.sig, &b->simple.sig_len,
	163	&b->simple.offset);
	164	}
	165
	166	static __be32
	167	nfsd4_block_proc_getdeviceinfo(struct super_block *sb,
d7c920d1	168	struct svc_rqst *rqstp,
f99d4fbd	169	struct nfs4_client *clp,
368248ee CH	170	struct nfsd4_getdeviceinfo *gdp)
	171	{
	172	if (sb->s_bdev != sb->s_bdev->bd_contains)
	173	return nfserr_inval;
	174	return nfserrno(nfsd4_block_get_device_info_simple(sb, gdp));
	175	}
	176
	177	static __be32
	178	nfsd4_block_proc_layoutcommit(struct inode *inode,
	179	struct nfsd4_layoutcommit *lcp)
	180	{
	181	struct iomap *iomaps;
	182	int nr_iomaps;
	183
	184	nr_iomaps = nfsd4_block_decode_layoutupdate(lcp->lc_up_layout,
93407472	185	lcp->lc_up_len, &iomaps, i_blocksize(inode));
368248ee CH	186	if (nr_iomaps < 0)
	187	return nfserrno(nr_iomaps);
	188
	189	return nfsd4_block_commit_blocks(inode, lcp, iomaps, nr_iomaps);
	190	}
	191
8650b8a0	192	const struct nfsd4_layout_ops bl_layout_ops = {
40cdc7a5 CH	193	/*
	194	* Pretend that we send notification to the client. This is a blatant
	195	* lie to force recent Linux clients to cache our device IDs.
	196	* We rarely ever change the device ID, so the harm of leaking deviceids
	197	* for a while isn't too bad. Unfortunately RFC5661 is a complete mess
	198	* in this regard, but I filed errata 4119 for this a while ago, and
	199	* hopefully the Linux client will eventually start caching deviceids
	200	* without this again.
	201	*/
	202	.notify_types =
	203	NOTIFY_DEVICEID4_DELETE \| NOTIFY_DEVICEID4_CHANGE,
8650b8a0 CH	204	.proc_getdeviceinfo = nfsd4_block_proc_getdeviceinfo,
	205	.encode_getdeviceinfo = nfsd4_block_encode_getdeviceinfo,
	206	.proc_layoutget = nfsd4_block_proc_layoutget,
	207	.encode_layoutget = nfsd4_block_encode_layoutget,
	208	.proc_layoutcommit = nfsd4_block_proc_layoutcommit,
	209	};
368248ee	210	#endif /* CONFIG_NFSD_BLOCKLAYOUT */
f99d4fbd CH	211
	212	#ifdef CONFIG_NFSD_SCSILAYOUT
	213	static int nfsd4_scsi_identify_device(struct block_device *bdev,
	214	struct pnfs_block_volume *b)
	215	{
	216	struct request_queue *q = bdev->bd_disk->queue;
	217	struct request *rq;
82ed4db4	218	struct scsi_request *req;
dac27072 SM	219	/*
	220	* The allocation length (passed in bytes 3 and 4 of the INQUIRY
	221	* command descriptor block) specifies the number of bytes that have
	222	* been allocated for the data-in buffer.
	223	* 252 is the highest one-byte value that is a multiple of 4.
	224	* 65532 is the highest two-byte value that is a multiple of 4.
	225	*/
	226	size_t bufflen = 252, maxlen = 65532, len, id_len;
f99d4fbd	227	u8 buf, d, type, assoc;
dac27072	228	int retries = 1, error;
f99d4fbd	229
30181faa BVA	230	if (WARN_ON_ONCE(!blk_queue_scsi_passthrough(q)))
	231	return -EINVAL;
	232
dac27072	233	again:
f99d4fbd CH	234	buf = kzalloc(bufflen, GFP_KERNEL);
	235	if (!buf)
	236	return -ENOMEM;
	237
ff005a06	238	rq = blk_get_request(q, REQ_OP_SCSI_IN, 0);
f99d4fbd CH	239	if (IS_ERR(rq)) {
	240	error = -ENOMEM;
	241	goto out_free_buf;
	242	}
82ed4db4	243	req = scsi_req(rq);
f99d4fbd CH	244
	245	error = blk_rq_map_kern(q, rq, buf, bufflen, GFP_KERNEL);
	246	if (error)
	247	goto out_put_request;
	248
82ed4db4 CH	249	req->cmd[0] = INQUIRY;
	250	req->cmd[1] = 1;
	251	req->cmd[2] = 0x83;
	252	req->cmd[3] = bufflen >> 8;
	253	req->cmd[4] = bufflen & 0xff;
	254	req->cmd_len = COMMAND_SIZE(INQUIRY);
f99d4fbd	255
b7819b92	256	blk_execute_rq(rq->q, NULL, rq, 1);
17d5363b	257	if (req->result) {
f99d4fbd	258	pr_err("pNFS: INQUIRY 0x83 failed with: %x\n",
17d5363b	259	req->result);
b7819b92	260	error = -EIO;
f99d4fbd CH	261	goto out_put_request;
	262	}
	263
	264	len = (buf[2] << 8) + buf[3] + 4;
	265	if (len > bufflen) {
dac27072 SM	266	if (len <= maxlen && retries--) {
	267	blk_put_request(rq);
	268	kfree(buf);
	269	bufflen = len;
	270	goto again;
	271	}
f99d4fbd CH	272	pr_err("pNFS: INQUIRY 0x83 response invalid (len = %zd)\n",
	273	len);
	274	goto out_put_request;
	275	}
	276
	277	d = buf + 4;
	278	for (d = buf + 4; d < buf + len; d += id_len + 4) {
	279	id_len = d[3];
	280	type = d[1] & 0xf;
	281	assoc = (d[1] >> 4) & 0x3;
	282
	283	/*
	284	* We only care about a EUI-64 and NAA designator types
	285	* with LU association.
	286	*/
	287	if (assoc != 0x00)
	288	continue;
	289	if (type != 0x02 && type != 0x03)
	290	continue;
	291	if (id_len != 8 && id_len != 12 && id_len != 16)
	292	continue;
	293
	294	b->scsi.code_set = PS_CODE_SET_BINARY;
	295	b->scsi.designator_type = type == 0x02 ?
	296	PS_DESIGNATOR_EUI64 : PS_DESIGNATOR_NAA;
	297	b->scsi.designator_len = id_len;
	298	memcpy(b->scsi.designator, d + 4, id_len);
	299
	300	/*
	301	* If we found a 8 or 12 byte descriptor continue on to
	302	* see if a 16 byte one is available. If we find a
	303	* 16 byte descriptor we're done.
	304	*/
	305	if (id_len == 16)
	306	break;
	307	}
	308
	309	out_put_request:
	310	blk_put_request(rq);
	311	out_free_buf:
	312	kfree(buf);
	313	return error;
	314	}
	315
eee93016	316	#define NFSD_MDS_PR_KEY 0x0100000000000000ULL
f99d4fbd CH	317
	318	/*
	319	* We use the client ID as a unique key for the reservations.
	320	* This allows us to easily fence a client when recalls fail.
	321	*/
	322	static u64 nfsd4_scsi_pr_key(struct nfs4_client *clp)
	323	{
	324	return ((u64)clp->cl_clientid.cl_boot << 32) \| clp->cl_clientid.cl_id;
	325	}
	326
	327	static int
	328	nfsd4_block_get_device_info_scsi(struct super_block *sb,
	329	struct nfs4_client *clp,
	330	struct nfsd4_getdeviceinfo *gdp)
	331	{
	332	struct pnfs_block_deviceaddr *dev;
	333	struct pnfs_block_volume *b;
	334	const struct pr_ops *ops;
	335	int error;
	336
	337	dev = kzalloc(sizeof(struct pnfs_block_deviceaddr) +
	338	sizeof(struct pnfs_block_volume), GFP_KERNEL);
	339	if (!dev)
	340	return -ENOMEM;
	341	gdp->gd_device = dev;
	342
	343	dev->nr_volumes = 1;
	344	b = &dev->volumes[0];
	345
	346	b->type = PNFS_BLOCK_VOLUME_SCSI;
	347	b->scsi.pr_key = nfsd4_scsi_pr_key(clp);
	348
	349	error = nfsd4_scsi_identify_device(sb->s_bdev, b);
	350	if (error)
	351	return error;
	352
	353	ops = sb->s_bdev->bd_disk->fops->pr_ops;
	354	if (!ops) {
	355	pr_err("pNFS: device %s does not support PRs.\n",
	356	sb->s_id);
	357	return -EINVAL;
	358	}
	359
	360	error = ops->pr_register(sb->s_bdev, 0, NFSD_MDS_PR_KEY, true);
	361	if (error) {
	362	pr_err("pNFS: failed to register key for device %s.\n",
	363	sb->s_id);
	364	return -EINVAL;
	365	}
	366
	367	error = ops->pr_reserve(sb->s_bdev, NFSD_MDS_PR_KEY,
	368	PR_EXCLUSIVE_ACCESS_REG_ONLY, 0);
	369	if (error) {
	370	pr_err("pNFS: failed to reserve device %s.\n",
	371	sb->s_id);
	372	return -EINVAL;
	373	}
	374
	375	return 0;
	376	}
	377
	378	static __be32
	379	nfsd4_scsi_proc_getdeviceinfo(struct super_block *sb,
d7c920d1	380	struct svc_rqst *rqstp,
f99d4fbd CH	381	struct nfs4_client *clp,
	382	struct nfsd4_getdeviceinfo *gdp)
	383	{
	384	if (sb->s_bdev != sb->s_bdev->bd_contains)
	385	return nfserr_inval;
	386	return nfserrno(nfsd4_block_get_device_info_scsi(sb, clp, gdp));
	387	}
	388	static __be32
	389	nfsd4_scsi_proc_layoutcommit(struct inode *inode,
	390	struct nfsd4_layoutcommit *lcp)
	391	{
	392	struct iomap *iomaps;
	393	int nr_iomaps;
	394
	395	nr_iomaps = nfsd4_scsi_decode_layoutupdate(lcp->lc_up_layout,
93407472	396	lcp->lc_up_len, &iomaps, i_blocksize(inode));
f99d4fbd CH	397	if (nr_iomaps < 0)
	398	return nfserrno(nr_iomaps);
	399
	400	return nfsd4_block_commit_blocks(inode, lcp, iomaps, nr_iomaps);
	401	}
	402
	403	static void
	404	nfsd4_scsi_fence_client(struct nfs4_layout_stateid *ls)
	405	{
	406	struct nfs4_client *clp = ls->ls_stid.sc_client;
	407	struct block_device *bdev = ls->ls_file->f_path.mnt->mnt_sb->s_bdev;
	408
	409	bdev->bd_disk->fops->pr_ops->pr_preempt(bdev, NFSD_MDS_PR_KEY,
	410	nfsd4_scsi_pr_key(clp), 0, true);
	411	}
	412
	413	const struct nfsd4_layout_ops scsi_layout_ops = {
	414	/*
	415	* Pretend that we send notification to the client. This is a blatant
	416	* lie to force recent Linux clients to cache our device IDs.
	417	* We rarely ever change the device ID, so the harm of leaking deviceids
	418	* for a while isn't too bad. Unfortunately RFC5661 is a complete mess
	419	* in this regard, but I filed errata 4119 for this a while ago, and
	420	* hopefully the Linux client will eventually start caching deviceids
	421	* without this again.
	422	*/
	423	.notify_types =
	424	NOTIFY_DEVICEID4_DELETE \| NOTIFY_DEVICEID4_CHANGE,
	425	.proc_getdeviceinfo = nfsd4_scsi_proc_getdeviceinfo,
	426	.encode_getdeviceinfo = nfsd4_block_encode_getdeviceinfo,
	427	.proc_layoutget = nfsd4_block_proc_layoutget,
	428	.encode_layoutget = nfsd4_block_encode_layoutget,
	429	.proc_layoutcommit = nfsd4_scsi_proc_layoutcommit,
	430	.fence_client = nfsd4_scsi_fence_client,
	431	};
	432	#endif /* CONFIG_NFSD_SCSILAYOUT */