This series adds support the pNFS SCSI layout to both the NFS server
and client. It's a fairly simple extension to the existing block
layout drivers, which relies on the Persistent Reservation API in
the block layer merged a while ago.
On the server side we will now always export SCSI devices a SCSI
layout. I though about allowing exports of multiple layour types
for the same file system, but both the not really production ready
nature of block layout fencing, and the horrors of passing options
to NFS exports made me go with the simple way for now.
Based on draft-ietf-nfsv4-scsi-layout-05 after the WG last call.
Signed-off-by: Christoph Hellwig <[email protected]>
---
include/linux/nfs4.h | 19 +++++++++++++++++++
1 file changed, 19 insertions(+)
diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h
index d6f9b4e..0114334 100644
--- a/include/linux/nfs4.h
+++ b/include/linux/nfs4.h
@@ -529,6 +529,7 @@ enum pnfs_layouttype {
LAYOUT_OSD2_OBJECTS = 2,
LAYOUT_BLOCK_VOLUME = 3,
LAYOUT_FLEX_FILES = 4,
+ LAYOUT_SCSI = 5,
LAYOUT_TYPE_MAX
};
@@ -555,6 +556,7 @@ enum pnfs_block_volume_type {
PNFS_BLOCK_VOLUME_SLICE = 1,
PNFS_BLOCK_VOLUME_CONCAT = 2,
PNFS_BLOCK_VOLUME_STRIPE = 3,
+ PNFS_BLOCK_VOLUME_SCSI = 4,
};
enum pnfs_block_extent_state {
@@ -568,6 +570,23 @@ enum pnfs_block_extent_state {
#define PNFS_BLOCK_EXTENT_SIZE \
(7 * sizeof(__be32) + NFS4_DEVICEID4_SIZE)
+/* on the wire size of a scsi commit range */
+#define PNFS_SCSI_RANGE_SIZE \
+ (4 * sizeof(__be32))
+
+enum scsi_code_set {
+ PS_CODE_SET_BINARY = 1,
+ PS_CODE_SET_ASCII = 2,
+ PS_CODE_SET_UTF8 = 3
+};
+
+enum scsi_designator_type {
+ PS_DESIGNATOR_T10 = 1,
+ PS_DESIGNATOR_EUI64 = 2,
+ PS_DESIGNATOR_NAA = 3,
+ PS_DESIGNATOR_NAME = 8
+};
+
#define NFL4_UFLG_MASK 0x0000003F
#define NFL4_UFLG_DENSE 0x00000001
#define NFL4_UFLG_COMMIT_THRU_MDS 0x00000002
--
2.1.4
This is a trivial extension to the block layout driver to support the
new SCSI layouts draft. There are three changes:
- device identifcation through the SCSI VPD page. This allows us to
directly use the udev generated persistent device names instead of
requiring an expensive lookup by crawling every block device node
in /dev and reading a signature for it.
- use of SCSI persistent reservations to protect device access and
allow for robust fencing. On the client sides this just means
registering and unregistering a server supplied key.
- an optimized LAYOUTCOMMIT payload that doesn't send unessecary
fields to the server.
Signed-off-by: Christoph Hellwig <[email protected]>
---
fs/nfs/blocklayout/blocklayout.c | 59 ++++++++++++++--
fs/nfs/blocklayout/blocklayout.h | 12 +++-
fs/nfs/blocklayout/dev.c | 144 ++++++++++++++++++++++++++++++++++++++-
fs/nfs/blocklayout/extent_tree.c | 44 ++++++++----
4 files changed, 236 insertions(+), 23 deletions(-)
diff --git a/fs/nfs/blocklayout/blocklayout.c b/fs/nfs/blocklayout/blocklayout.c
index ddd0138..b27c409 100644
--- a/fs/nfs/blocklayout/blocklayout.c
+++ b/fs/nfs/blocklayout/blocklayout.c
@@ -446,8 +446,8 @@ static void bl_free_layout_hdr(struct pnfs_layout_hdr *lo)
kfree(bl);
}
-static struct pnfs_layout_hdr *bl_alloc_layout_hdr(struct inode *inode,
- gfp_t gfp_flags)
+static struct pnfs_layout_hdr *__bl_alloc_layout_hdr(struct inode *inode,
+ gfp_t gfp_flags, bool is_scsi_layout)
{
struct pnfs_block_layout *bl;
@@ -460,9 +460,22 @@ static struct pnfs_layout_hdr *bl_alloc_layout_hdr(struct inode *inode,
bl->bl_ext_ro = RB_ROOT;
spin_lock_init(&bl->bl_ext_lock);
+ bl->bl_scsi_layout = is_scsi_layout;
return &bl->bl_layout;
}
+static struct pnfs_layout_hdr *bl_alloc_layout_hdr(struct inode *inode,
+ gfp_t gfp_flags)
+{
+ return __bl_alloc_layout_hdr(inode, gfp_flags, false);
+}
+
+static struct pnfs_layout_hdr *sl_alloc_layout_hdr(struct inode *inode,
+ gfp_t gfp_flags)
+{
+ return __bl_alloc_layout_hdr(inode, gfp_flags, true);
+}
+
static void bl_free_lseg(struct pnfs_layout_segment *lseg)
{
dprintk("%s enter\n", __func__);
@@ -888,22 +901,53 @@ static struct pnfs_layoutdriver_type blocklayout_type = {
.sync = pnfs_generic_sync,
};
+static struct pnfs_layoutdriver_type scsilayout_type = {
+ .id = LAYOUT_SCSI,
+ .name = "LAYOUT_SCSI",
+ .owner = THIS_MODULE,
+ .flags = PNFS_LAYOUTRET_ON_SETATTR |
+ PNFS_READ_WHOLE_PAGE,
+ .read_pagelist = bl_read_pagelist,
+ .write_pagelist = bl_write_pagelist,
+ .alloc_layout_hdr = sl_alloc_layout_hdr,
+ .free_layout_hdr = bl_free_layout_hdr,
+ .alloc_lseg = bl_alloc_lseg,
+ .free_lseg = bl_free_lseg,
+ .return_range = bl_return_range,
+ .prepare_layoutcommit = bl_prepare_layoutcommit,
+ .cleanup_layoutcommit = bl_cleanup_layoutcommit,
+ .set_layoutdriver = bl_set_layoutdriver,
+ .alloc_deviceid_node = bl_alloc_deviceid_node,
+ .free_deviceid_node = bl_free_deviceid_node,
+ .pg_read_ops = &bl_pg_read_ops,
+ .pg_write_ops = &bl_pg_write_ops,
+ .sync = pnfs_generic_sync,
+};
+
+
static int __init nfs4blocklayout_init(void)
{
int ret;
dprintk("%s: NFSv4 Block Layout Driver Registering...\n", __func__);
- ret = pnfs_register_layoutdriver(&blocklayout_type);
+ ret = bl_init_pipefs();
if (ret)
goto out;
- ret = bl_init_pipefs();
+
+ ret = pnfs_register_layoutdriver(&blocklayout_type);
if (ret)
- goto out_unregister;
+ goto out_cleanup_pipe;
+
+ ret = pnfs_register_layoutdriver(&scsilayout_type);
+ if (ret)
+ goto out_unregister_block;
return 0;
-out_unregister:
+out_unregister_block:
pnfs_unregister_layoutdriver(&blocklayout_type);
+out_cleanup_pipe:
+ bl_cleanup_pipefs();
out:
return ret;
}
@@ -913,8 +957,9 @@ static void __exit nfs4blocklayout_exit(void)
dprintk("%s: NFSv4 Block Layout Driver Unregistering...\n",
__func__);
- bl_cleanup_pipefs();
+ pnfs_unregister_layoutdriver(&scsilayout_type);
pnfs_unregister_layoutdriver(&blocklayout_type);
+ bl_cleanup_pipefs();
}
MODULE_ALIAS("nfs-layouttype4-3");
diff --git a/fs/nfs/blocklayout/blocklayout.h b/fs/nfs/blocklayout/blocklayout.h
index c556640..6e9b470 100644
--- a/fs/nfs/blocklayout/blocklayout.h
+++ b/fs/nfs/blocklayout/blocklayout.h
@@ -55,7 +55,6 @@ struct pnfs_block_dev;
*/
#define PNFS_BLOCK_UUID_LEN 128
-
struct pnfs_block_volume {
enum pnfs_block_volume_type type;
union {
@@ -82,6 +81,13 @@ struct pnfs_block_volume {
u32 volumes_count;
u32 volumes[PNFS_BLOCK_MAX_DEVICES];
} stripe;
+ struct {
+ enum scsi_code_set code_set;
+ enum scsi_designator_type designator_type;
+ int designator_len;
+ u8 designator[256];
+ u64 pr_key;
+ } scsi;
};
};
@@ -106,6 +112,9 @@ struct pnfs_block_dev {
struct block_device *bdev;
u64 disk_offset;
+ u64 pr_key;
+ bool pr_registered;
+
bool (*map)(struct pnfs_block_dev *dev, u64 offset,
struct pnfs_block_dev_map *map);
};
@@ -131,6 +140,7 @@ struct pnfs_block_layout {
struct rb_root bl_ext_rw;
struct rb_root bl_ext_ro;
spinlock_t bl_ext_lock; /* Protects list manipulation */
+ bool bl_scsi_layout;
};
static inline struct pnfs_block_layout *
diff --git a/fs/nfs/blocklayout/dev.c b/fs/nfs/blocklayout/dev.c
index a861bbd..31b0f6b 100644
--- a/fs/nfs/blocklayout/dev.c
+++ b/fs/nfs/blocklayout/dev.c
@@ -1,11 +1,12 @@
/*
- * Copyright (c) 2014 Christoph Hellwig.
+ * Copyright (c) 2014-2016 Christoph Hellwig.
*/
#include <linux/sunrpc/svc.h>
#include <linux/blkdev.h>
#include <linux/nfs4.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_xdr.h>
+#include <linux/pr.h>
#include "blocklayout.h"
@@ -21,6 +22,17 @@ bl_free_device(struct pnfs_block_dev *dev)
bl_free_device(&dev->children[i]);
kfree(dev->children);
} else {
+ if (dev->pr_registered) {
+ const struct pr_ops *ops =
+ dev->bdev->bd_disk->fops->pr_ops;
+ int error;
+
+ error = ops->pr_register(dev->bdev, dev->pr_key, 0,
+ false);
+ if (error)
+ pr_err("failed to unregister PR key.\n");
+ }
+
if (dev->bdev)
blkdev_put(dev->bdev, FMODE_READ | FMODE_WRITE);
}
@@ -113,6 +125,24 @@ nfs4_block_decode_volume(struct xdr_stream *xdr, struct pnfs_block_volume *b)
for (i = 0; i < b->stripe.volumes_count; i++)
b->stripe.volumes[i] = be32_to_cpup(p++);
break;
+ case PNFS_BLOCK_VOLUME_SCSI:
+ p = xdr_inline_decode(xdr, 4 + 4 + 4);
+ if (!p)
+ return -EIO;
+ b->scsi.code_set = be32_to_cpup(p++);
+ b->scsi.designator_type = be32_to_cpup(p++);
+ b->scsi.designator_len = be32_to_cpup(p++);
+ p = xdr_inline_decode(xdr, b->scsi.designator_len);
+ if (!p)
+ return -EIO;
+ if (b->scsi.designator_len > 256)
+ return -EIO;
+ memcpy(&b->scsi.designator, p, b->scsi.designator_len);
+ p = xdr_inline_decode(xdr, 8);
+ if (!p)
+ return -EIO;
+ p = xdr_decode_hyper(p, &b->scsi.pr_key);
+ break;
default:
dprintk("unknown volume type!\n");
return -EIO;
@@ -216,6 +246,116 @@ bl_parse_simple(struct nfs_server *server, struct pnfs_block_dev *d,
return 0;
}
+static bool
+bl_validate_designator(struct pnfs_block_volume *v)
+{
+ switch (v->scsi.designator_type) {
+ case PS_DESIGNATOR_EUI64:
+ if (v->scsi.code_set != PS_CODE_SET_BINARY)
+ return false;
+
+ if (v->scsi.designator_len != 8 &&
+ v->scsi.designator_len != 10 &&
+ v->scsi.designator_len != 16)
+ return false;
+
+ return true;
+ case PS_DESIGNATOR_NAA:
+ if (v->scsi.code_set != PS_CODE_SET_BINARY)
+ return false;
+
+ if (v->scsi.designator_len != 8 &&
+ v->scsi.designator_len != 16)
+ return false;
+
+ return true;
+ case PS_DESIGNATOR_T10:
+ case PS_DESIGNATOR_NAME:
+ pr_err("pNFS: unsupported designator "
+ "(code set %d, type %d, len %d.\n",
+ v->scsi.code_set,
+ v->scsi.designator_type,
+ v->scsi.designator_len);
+ return false;
+ default:
+ pr_err("pNFS: invalid designator "
+ "(code set %d, type %d, len %d.\n",
+ v->scsi.code_set,
+ v->scsi.designator_type,
+ v->scsi.designator_len);
+ return false;
+ }
+}
+
+static int
+bl_parse_scsi(struct nfs_server *server, struct pnfs_block_dev *d,
+ struct pnfs_block_volume *volumes, int idx, gfp_t gfp_mask)
+{
+ struct pnfs_block_volume *v = &volumes[idx];
+ const struct pr_ops *ops;
+ const char *devname;
+ int error;
+
+ if (!bl_validate_designator(v))
+ return -EINVAL;
+
+ switch (v->scsi.designator_len) {
+ case 8:
+ devname = kasprintf(GFP_KERNEL, "/dev/disk/by-id/wwn-0x%8phN",
+ v->scsi.designator);
+ break;
+ case 12:
+ devname = kasprintf(GFP_KERNEL, "/dev/disk/by-id/wwn-0x%12phN",
+ v->scsi.designator);
+ break;
+ case 16:
+ devname = kasprintf(GFP_KERNEL, "/dev/disk/by-id/wwn-0x%16phN",
+ v->scsi.designator);
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ d->bdev = blkdev_get_by_path(devname, FMODE_READ, NULL);
+ if (IS_ERR(d->bdev)) {
+ pr_warn("pNFS: failed to open device %s (%ld)\n",
+ devname, PTR_ERR(d->bdev));
+ kfree(devname);
+ return PTR_ERR(d->bdev);
+ }
+
+ kfree(devname);
+
+ d->len = i_size_read(d->bdev->bd_inode);
+ d->map = bl_map_simple;
+ d->pr_key = v->scsi.pr_key;
+
+ pr_info("pNFS: using block device %s (reservation key 0x%llx)\n",
+ d->bdev->bd_disk->disk_name, d->pr_key);
+
+ ops = d->bdev->bd_disk->fops->pr_ops;
+ if (!ops) {
+ pr_err("pNFS: block device %s does not support reservations.",
+ d->bdev->bd_disk->disk_name);
+ error = -EINVAL;
+ goto out_blkdev_put;
+ }
+
+ error = ops->pr_register(d->bdev, 0, d->pr_key, true);
+ if (error) {
+ pr_err("pNFS: failed to register key for block device %s.",
+ d->bdev->bd_disk->disk_name);
+ goto out_blkdev_put;
+ }
+
+ d->pr_registered = true;
+ return 0;
+
+out_blkdev_put:
+ blkdev_put(d->bdev, FMODE_READ);
+ return error;
+}
+
static int
bl_parse_slice(struct nfs_server *server, struct pnfs_block_dev *d,
struct pnfs_block_volume *volumes, int idx, gfp_t gfp_mask)
@@ -303,6 +443,8 @@ bl_parse_deviceid(struct nfs_server *server, struct pnfs_block_dev *d,
return bl_parse_concat(server, d, volumes, idx, gfp_mask);
case PNFS_BLOCK_VOLUME_STRIPE:
return bl_parse_stripe(server, d, volumes, idx, gfp_mask);
+ case PNFS_BLOCK_VOLUME_SCSI:
+ return bl_parse_scsi(server, d, volumes, idx, gfp_mask);
default:
dprintk("unsupported volume type: %d\n", volumes[idx].type);
return -EIO;
diff --git a/fs/nfs/blocklayout/extent_tree.c b/fs/nfs/blocklayout/extent_tree.c
index c59a59c..df366fc 100644
--- a/fs/nfs/blocklayout/extent_tree.c
+++ b/fs/nfs/blocklayout/extent_tree.c
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2014 Christoph Hellwig.
+ * Copyright (c) 2014-2016 Christoph Hellwig.
*/
#include <linux/vmalloc.h>
@@ -462,10 +462,12 @@ out:
return err;
}
-static size_t ext_tree_layoutupdate_size(size_t count)
+static size_t ext_tree_layoutupdate_size(struct pnfs_block_layout *bl, size_t count)
{
- return sizeof(__be32) /* number of entries */ +
- PNFS_BLOCK_EXTENT_SIZE * count;
+ if (bl->bl_scsi_layout)
+ return sizeof(__be32) + PNFS_SCSI_RANGE_SIZE * count;
+ else
+ return sizeof(__be32) + PNFS_BLOCK_EXTENT_SIZE * count;
}
static void ext_tree_free_commitdata(struct nfs4_layoutcommit_args *arg,
@@ -482,6 +484,23 @@ static void ext_tree_free_commitdata(struct nfs4_layoutcommit_args *arg,
}
}
+static __be32 *encode_block_extent(struct pnfs_block_extent *be, __be32 *p)
+{
+ p = xdr_encode_opaque_fixed(p, be->be_device->deviceid.data,
+ NFS4_DEVICEID4_SIZE);
+ p = xdr_encode_hyper(p, be->be_f_offset << SECTOR_SHIFT);
+ p = xdr_encode_hyper(p, be->be_length << SECTOR_SHIFT);
+ p = xdr_encode_hyper(p, 0LL);
+ *p++ = cpu_to_be32(PNFS_BLOCK_READWRITE_DATA);
+ return p;
+}
+
+static __be32 *encode_scsi_range(struct pnfs_block_extent *be, __be32 *p)
+{
+ p = xdr_encode_hyper(p, be->be_f_offset << SECTOR_SHIFT);
+ return xdr_encode_hyper(p, be->be_length << SECTOR_SHIFT);
+}
+
static int ext_tree_encode_commit(struct pnfs_block_layout *bl, __be32 *p,
size_t buffer_size, size_t *count)
{
@@ -495,19 +514,16 @@ static int ext_tree_encode_commit(struct pnfs_block_layout *bl, __be32 *p,
continue;
(*count)++;
- if (ext_tree_layoutupdate_size(*count) > buffer_size) {
+ if (ext_tree_layoutupdate_size(bl, *count) > buffer_size) {
/* keep counting.. */
ret = -ENOSPC;
continue;
}
- p = xdr_encode_opaque_fixed(p, be->be_device->deviceid.data,
- NFS4_DEVICEID4_SIZE);
- p = xdr_encode_hyper(p, be->be_f_offset << SECTOR_SHIFT);
- p = xdr_encode_hyper(p, be->be_length << SECTOR_SHIFT);
- p = xdr_encode_hyper(p, 0LL);
- *p++ = cpu_to_be32(PNFS_BLOCK_READWRITE_DATA);
-
+ if (bl->bl_scsi_layout)
+ p = encode_scsi_range(be, p);
+ else
+ p = encode_block_extent(be, p);
be->be_tag = EXTENT_COMMITTING;
}
spin_unlock(&bl->bl_ext_lock);
@@ -536,7 +552,7 @@ retry:
if (unlikely(ret)) {
ext_tree_free_commitdata(arg, buffer_size);
- buffer_size = ext_tree_layoutupdate_size(count);
+ buffer_size = ext_tree_layoutupdate_size(bl, count);
count = 0;
arg->layoutupdate_pages =
@@ -555,7 +571,7 @@ retry:
}
*start_p = cpu_to_be32(count);
- arg->layoutupdate_len = ext_tree_layoutupdate_size(count);
+ arg->layoutupdate_len = ext_tree_layoutupdate_size(bl, count);
if (unlikely(arg->layoutupdate_pages != &arg->layoutupdate_page)) {
void *p = start_p, *end = p + arg->layoutupdate_len;
--
2.1.4
This is a simple extension to the block layout driver to use SCSI
persistent reservations for access control and fencing, as well as
SCSI VPD pages for device identification.
For this we need to pass the nfs4_client to the proc_getdeviceinfo method
to generate the reservation key, and add a new fence_client method
to allow for fence actions in the layout driver.
Signed-off-by: Christoph Hellwig <[email protected]>
---
fs/nfsd/blocklayout.c | 234 +++++++++++++++++++++++++++++++++++++++++++++--
fs/nfsd/blocklayoutxdr.c | 65 ++++++++++++-
fs/nfsd/blocklayoutxdr.h | 14 +++
fs/nfsd/nfs4layouts.c | 28 ++++--
fs/nfsd/nfs4proc.c | 6 +-
fs/nfsd/pnfs.h | 4 +
6 files changed, 331 insertions(+), 20 deletions(-)
diff --git a/fs/nfsd/blocklayout.c b/fs/nfsd/blocklayout.c
index c29d942..48308ee 100644
--- a/fs/nfsd/blocklayout.c
+++ b/fs/nfsd/blocklayout.c
@@ -1,11 +1,14 @@
/*
- * Copyright (c) 2014 Christoph Hellwig.
+ * 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"
@@ -37,6 +40,7 @@ nfsd4_block_get_device_info_simple(struct super_block *sb,
static __be32
nfsd4_block_proc_getdeviceinfo(struct super_block *sb,
+ struct nfs4_client *clp,
struct nfsd4_getdeviceinfo *gdp)
{
if (sb->s_bdev != sb->s_bdev->bd_contains)
@@ -44,6 +48,163 @@ nfsd4_block_proc_getdeviceinfo(struct super_block *sb,
return nfserrno(nfsd4_block_get_device_info_simple(sb, gdp));
}
+
+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);
+
+ }
+
+ 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 0x0100000000000000
+
+/*
+ * We use the client ID as a uniqueue 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 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_block_proc_layoutget(struct inode *inode, const struct svc_fh *fhp,
struct nfsd4_layoutget *args)
@@ -141,20 +302,13 @@ out_layoutunavailable:
}
static __be32
-nfsd4_block_proc_layoutcommit(struct inode *inode,
- struct nfsd4_layoutcommit *lcp)
+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 };
- struct iomap *iomaps;
- int nr_iomaps;
int error;
- 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);
-
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);
@@ -172,6 +326,46 @@ nfsd4_block_proc_layoutcommit(struct inode *inode,
return nfserrno(error);
}
+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);
+}
+
+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 bl_layout_ops = {
/*
* Pretend that we send notification to the client. This is a blatant
@@ -190,3 +384,23 @@ const struct nfsd4_layout_ops bl_layout_ops = {
.encode_layoutget = nfsd4_block_encode_layoutget,
.proc_layoutcommit = nfsd4_block_proc_layoutcommit,
};
+
+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,
+};
diff --git a/fs/nfsd/blocklayoutxdr.c b/fs/nfsd/blocklayoutxdr.c
index 6d834dc..ca18836 100644
--- a/fs/nfsd/blocklayoutxdr.c
+++ b/fs/nfsd/blocklayoutxdr.c
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2014 Christoph Hellwig.
+ * Copyright (c) 2014-2016 Christoph Hellwig.
*/
#include <linux/sunrpc/svc.h>
#include <linux/exportfs.h>
@@ -53,6 +53,18 @@ nfsd4_block_encode_volume(struct xdr_stream *xdr, struct pnfs_block_volume *b)
p = xdr_encode_hyper(p, b->simple.offset);
p = xdr_encode_opaque(p, b->simple.sig, b->simple.sig_len);
break;
+ case PNFS_BLOCK_VOLUME_SCSI:
+ len = 4 + 4 + 4 + 4 + b->scsi.designator_len + 8;
+ p = xdr_reserve_space(xdr, len);
+ if (!p)
+ return -ETOOSMALL;
+
+ *p++ = cpu_to_be32(b->type);
+ *p++ = cpu_to_be32(b->scsi.code_set);
+ *p++ = cpu_to_be32(b->scsi.designator_type);
+ p = xdr_encode_opaque(p, b->scsi.designator, b->scsi.designator_len);
+ p = xdr_encode_hyper(p, b->scsi.pr_key);
+ break;
default:
return -ENOTSUPP;
}
@@ -155,3 +167,54 @@ fail:
kfree(iomaps);
return -EINVAL;
}
+
+int
+nfsd4_scsi_decode_layoutupdate(__be32 *p, u32 len, struct iomap **iomapp,
+ u32 block_size)
+{
+ struct iomap *iomaps;
+ u32 nr_iomaps, expected, i;
+
+ if (len < sizeof(u32)) {
+ dprintk("%s: extent array too small: %u\n", __func__, len);
+ return -EINVAL;
+ }
+
+ nr_iomaps = be32_to_cpup(p++);
+ expected = sizeof(__be32) + nr_iomaps * PNFS_SCSI_RANGE_SIZE;
+ if (len != expected) {
+ dprintk("%s: extent array size mismatch: %u/%u\n",
+ __func__, len, expected);
+ return -EINVAL;
+ }
+
+ iomaps = kcalloc(nr_iomaps, sizeof(*iomaps), GFP_KERNEL);
+ if (!iomaps) {
+ dprintk("%s: failed to allocate extent array\n", __func__);
+ return -ENOMEM;
+ }
+
+ for (i = 0; i < nr_iomaps; i++) {
+ u64 val;
+
+ p = xdr_decode_hyper(p, &val);
+ if (val & (block_size - 1)) {
+ dprintk("%s: unaligned offset 0x%llx\n", __func__, val);
+ goto fail;
+ }
+ iomaps[i].offset = val;
+
+ p = xdr_decode_hyper(p, &val);
+ if (val & (block_size - 1)) {
+ dprintk("%s: unaligned length 0x%llx\n", __func__, val);
+ goto fail;
+ }
+ iomaps[i].length = val;
+ }
+
+ *iomapp = iomaps;
+ return nr_iomaps;
+fail:
+ kfree(iomaps);
+ return -EINVAL;
+}
diff --git a/fs/nfsd/blocklayoutxdr.h b/fs/nfsd/blocklayoutxdr.h
index 6de925f..397bc75 100644
--- a/fs/nfsd/blocklayoutxdr.h
+++ b/fs/nfsd/blocklayoutxdr.h
@@ -15,6 +15,11 @@ struct pnfs_block_extent {
enum pnfs_block_extent_state es;
};
+struct pnfs_block_range {
+ u64 foff;
+ u64 len;
+};
+
/*
* Random upper cap for the uuid length to avoid unbounded allocation.
* Not actually limited by the protocol.
@@ -29,6 +34,13 @@ struct pnfs_block_volume {
u32 sig_len;
u8 sig[PNFS_BLOCK_UUID_LEN];
} simple;
+ struct {
+ enum scsi_code_set code_set;
+ enum scsi_designator_type designator_type;
+ int designator_len;
+ u8 designator[256];
+ u64 pr_key;
+ } scsi;
};
};
@@ -43,5 +55,7 @@ __be32 nfsd4_block_encode_layoutget(struct xdr_stream *xdr,
struct nfsd4_layoutget *lgp);
int nfsd4_block_decode_layoutupdate(__be32 *p, u32 len, struct iomap **iomapp,
u32 block_size);
+int nfsd4_scsi_decode_layoutupdate(__be32 *p, u32 len, struct iomap **iomapp,
+ u32 block_size);
#endif /* _NFSD_BLOCKLAYOUTXDR_H */
diff --git a/fs/nfsd/nfs4layouts.c b/fs/nfsd/nfs4layouts.c
index ce2d010..786e59e 100644
--- a/fs/nfsd/nfs4layouts.c
+++ b/fs/nfsd/nfs4layouts.c
@@ -1,6 +1,7 @@
/*
* Copyright (c) 2014 Christoph Hellwig.
*/
+#include <linux/blkdev.h>
#include <linux/kmod.h>
#include <linux/file.h>
#include <linux/jhash.h>
@@ -27,6 +28,7 @@ static const struct lock_manager_operations nfsd4_layouts_lm_ops;
const struct nfsd4_layout_ops *nfsd4_layout_ops[LAYOUT_TYPE_MAX] = {
[LAYOUT_BLOCK_VOLUME] = &bl_layout_ops,
+ [LAYOUT_SCSI] = &scsi_layout_ops,
};
/* pNFS device ID to export fsid mapping */
@@ -121,10 +123,17 @@ void nfsd4_setup_layout_type(struct svc_export *exp)
if (!(exp->ex_flags & NFSEXP_PNFS))
return;
- if (sb->s_export_op->get_uuid &&
- sb->s_export_op->map_blocks &&
- sb->s_export_op->commit_blocks)
- exp->ex_layout_type = LAYOUT_BLOCK_VOLUME;
+ /*
+ * Check if the file systems supports exporting a block-like layout.
+ * If the block device supports reservations prefer the SCSI layout,
+ * else advertise the block layout.
+ */
+ if (sb->s_export_op->map_blocks && sb->s_export_op->commit_blocks) {
+ if (sb->s_bdev && sb->s_bdev->bd_disk->fops->pr_ops)
+ exp->ex_layout_type = LAYOUT_SCSI;
+ else if (sb->s_export_op->get_uuid)
+ exp->ex_layout_type = LAYOUT_BLOCK_VOLUME;
+ }
}
static void
@@ -590,8 +599,6 @@ nfsd4_cb_layout_fail(struct nfs4_layout_stateid *ls)
rpc_ntop((struct sockaddr *)&clp->cl_addr, addr_str, sizeof(addr_str));
- trace_layout_recall_fail(&ls->ls_stid.sc_stateid);
-
printk(KERN_WARNING
"nfsd: client %s failed to respond to layout recall. "
" Fencing..\n", addr_str);
@@ -626,6 +633,7 @@ nfsd4_cb_layout_done(struct nfsd4_callback *cb, struct rpc_task *task)
container_of(cb, struct nfs4_layout_stateid, ls_recall);
struct nfsd_net *nn;
ktime_t now, cutoff;
+ const struct nfsd4_layout_ops *ops;
LIST_HEAD(reaplist);
@@ -661,7 +669,13 @@ nfsd4_cb_layout_done(struct nfsd4_callback *cb, struct rpc_task *task)
/*
* Unknown error or non-responding client, we'll need to fence.
*/
- nfsd4_cb_layout_fail(ls);
+ trace_layout_recall_fail(&ls->ls_stid.sc_stateid);
+
+ ops = nfsd4_layout_ops[ls->ls_layout_type];
+ if (ops->fence_client)
+ ops->fence_client(ls);
+ else
+ nfsd4_cb_layout_fail(ls);
return -1;
}
}
diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
index 4cba786..629443e 100644
--- a/fs/nfsd/nfs4proc.c
+++ b/fs/nfsd/nfs4proc.c
@@ -1269,8 +1269,10 @@ nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
goto out;
nfserr = nfs_ok;
- if (gdp->gd_maxcount != 0)
- nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb, gdp);
+ if (gdp->gd_maxcount != 0) {
+ nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
+ cstate->session->se_client, gdp);
+ }
gdp->gd_notify_types &= ops->notify_types;
out:
diff --git a/fs/nfsd/pnfs.h b/fs/nfsd/pnfs.h
index d4c4453..3b54e43 100644
--- a/fs/nfsd/pnfs.h
+++ b/fs/nfsd/pnfs.h
@@ -21,6 +21,7 @@ struct nfsd4_layout_ops {
u32 notify_types;
__be32 (*proc_getdeviceinfo)(struct super_block *sb,
+ struct nfs4_client *clp,
struct nfsd4_getdeviceinfo *gdevp);
__be32 (*encode_getdeviceinfo)(struct xdr_stream *xdr,
struct nfsd4_getdeviceinfo *gdevp);
@@ -32,10 +33,13 @@ struct nfsd4_layout_ops {
__be32 (*proc_layoutcommit)(struct inode *inode,
struct nfsd4_layoutcommit *lcp);
+
+ void (*fence_client)(struct nfs4_layout_stateid *ls);
};
extern const struct nfsd4_layout_ops *nfsd4_layout_ops[];
extern const struct nfsd4_layout_ops bl_layout_ops;
+extern const struct nfsd4_layout_ops scsi_layout_ops;
__be32 nfsd4_preprocess_layout_stateid(struct svc_rqst *rqstp,
struct nfsd4_compound_state *cstate, stateid_t *stateid,
--
2.1.4
Hi Christoph,
[auto build test WARNING on nfsd/nfsd-next]
[also build test WARNING on v4.5-rc6 next-20160229]
[if your patch is applied to the wrong git tree, please drop us a note to help improving the system]
url: https://github.com/0day-ci/linux/commits/Christoph-Hellwig/nfs4-h-add-SCSI-layout-defintions/20160229-212616
base: git://linux-nfs.org/~bfields/linux.git nfsd-next
config: blackfin-allyesconfig (attached as .config)
reproduce:
wget https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/plain/sbin/make.cross -O ~/bin/make.cross
chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
make.cross ARCH=blackfin
All warnings (new ones prefixed by >>):
>> WARNING: fs/nfs/built-in.o(.init.text+0x2c20): Section mismatch in reference from the function _nfs4blocklayout_init() to the function .exit.text:_bl_cleanup_pipefs()
The function __init _nfs4blocklayout_init() references
a function __exit _bl_cleanup_pipefs().
This is often seen when error handling in the init function
uses functionality in the exit path.
The fix is often to remove the __exit annotation of
_bl_cleanup_pipefs() so it may be used outside an exit section.
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
Hi Christoph,
[auto build test WARNING on nfsd/nfsd-next]
[also build test WARNING on v4.5-rc6 next-20160229]
[if your patch is applied to the wrong git tree, please drop us a note to help improving the system]
url: https://github.com/0day-ci/linux/commits/Christoph-Hellwig/nfs4-h-add-SCSI-layout-defintions/20160229-212616
base: git://linux-nfs.org/~bfields/linux.git nfsd-next
config: mn10300-allmodconfig (attached as .config)
reproduce:
wget https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/plain/sbin/make.cross -O ~/bin/make.cross
chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
make.cross ARCH=mn10300
All warnings (new ones prefixed by >>):
>> WARNING: fs/nfs/blocklayout/blocklayoutdriver.o(.init.text+0x67): Section mismatch in reference from the variable .L302 to the function .exit.text:bl_cleanup_pipefs()
The variable __init .L302 references
a function __exit bl_cleanup_pipefs().
This is often seen when error handling in the init function
uses functionality in the exit path.
The fix is often to remove the __exit annotation of
bl_cleanup_pipefs() so it may be used outside an exit section.
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
On Mon, Feb 29, 2016 at 02:24:04PM +0100, Christoph Hellwig wrote:
> This series adds support the pNFS SCSI layout to both the NFS server
> and client. It's a fairly simple extension to the existing block
> layout drivers, which relies on the Persistent Reservation API in
> the block layer merged a while ago.
Thanks, Christoph!
> On the server side we will now always export SCSI devices a SCSI
> layout.
If I understand correctly, you still require NFSEXP_PNFS (the "pnfs"
export option). So, setting exporting with the "pnfs" option means
clients will see only one of block or SCSI layout support (or neither)
depending on what's possible, with preference for the SCSI layout.
> I though about allowing exports of multiple layour types
> for the same file system, but both the not really production ready
> nature of block layout fencing, and the horrors of passing options
> to NFS exports made me go with the simple way for now.
I agree that we should keep things simple, and agree that we should not
worry too much about block layout support.
I still wonder whether this is the best behavior. How about also adding
the ability to configure out the block layout? I think we'd want to be
sure it's completely off in production.
Also todo?:
- wireshark support
- testing, including of fencing and reboot recovery. (And:
currently I just share a file between two vm's to use as my
shared block device, I guess I'll need to set up a real SCSI
target instead.)
I think there's not any additional documentation required--this is just
like block layout but without the need to configure fencing. (Or do
people need to do some additional configuration of the SCSI devices? Or
check the specs on their hardware to make sure it has support for the
right features?)
--b.
On Tue, Mar 01, 2016 at 02:24:08PM -0500, J. Bruce Fields wrote:
> If I understand correctly, you still require NFSEXP_PNFS (the "pnfs"
> export option). So, setting exporting with the "pnfs" option means
> clients will see only one of block or SCSI layout support (or neither)
> depending on what's possible, with preference for the SCSI layout.
pnfs is supposed to enable any pnfs layout. Currently only block
is implemented, and with this patch the scsi layout as well.
The scsi layout will be exported if the device supports reservations
and the file system supports block layout exports (SCSI layout is the
same to the file system). Otherwise the block layout is enabled if
we're on a block device, and the file system supports block layout
exports.
> I still wonder whether this is the best behavior. How about also adding
> the ability to configure out the block layout? I think we'd want to be
> sure it's completely off in production.
Being able to configure block layout eventually might be useful. Do
you simply want separate compile time options? It's not like a whole
lot of code will go away, it's mostly just entry points.
> Also todo?:
> - wireshark support
I've mostly stayed clear of wiresharė.
> - testing, including of fencing and reboot recovery.
I've done a fair amount of testing, and the interesting part is
indeed fencing. The normal path is not very different from the block
layout driver.
> (And:
> currently I just share a file between two vm's to use as my
> shared block device, I guess I'll need to set up a real SCSI
> target instead.)
The in-kernel lio target will work just fine. You can actually export
that directly to VMs using the vhost_scsi driver, but I've not tried
that option myself yet.
> I think there's not any additional documentation required--this is just
> like block layout but without the need to configure fencing.
Exactly.
> (Or do
> people need to do some additional configuration of the SCSI devices? Or
> check the specs on their hardware to make sure it has support for the
> right features?)
Yes, the device needs to support persistent reservations.
I can write up a little blurb on that.
On Tue, Mar 01, 2016 at 09:30:12PM +0100, Christoph Hellwig wrote:
> On Tue, Mar 01, 2016 at 02:24:08PM -0500, J. Bruce Fields wrote:
> > I still wonder whether this is the best behavior. How about also adding
> > the ability to configure out the block layout? I think we'd want to be
> > sure it's completely off in production.
>
> Being able to configure block layout eventually might be useful. Do
> you simply want separate compile time options? It's not like a whole
> lot of code will go away, it's mostly just entry points.
Yeah, I wasn't thinking about the size of the code so much as
supportability.
> > Also todo?:
> > - wireshark support
>
> I've mostly stayed clear of wiresharė.
No problem, just noting it so I don't forget.
> > - testing, including of fencing and reboot recovery.
>
> I've done a fair amount of testing, and the interesting part is
> indeed fencing. The normal path is not very different from the block
> layout driver.
>
> > (And:
> > currently I just share a file between two vm's to use as my
> > shared block device, I guess I'll need to set up a real SCSI
> > target instead.)
>
> The in-kernel lio target will work just fine. You can actually export
> that directly to VMs using the vhost_scsi driver, but I've not tried
> that option myself yet.
>
> > I think there's not any additional documentation required--this is just
> > like block layout but without the need to configure fencing.
>
> Exactly.
>
> > (Or do
> > people need to do some additional configuration of the SCSI devices? Or
> > check the specs on their hardware to make sure it has support for the
> > right features?)
>
> Yes, the device needs to support persistent reservations.
>
> I can write up a little blurb on that.
OK, thanks.
--b.