From: Olga Kornievskaia <[email protected]>
This patch series adds client-side support for doing NFSv4.2 "inter"
copy offload between different NFS servers.
In case of the "inter" SSC copy files reside on different servers and
thus under different superblocks and require that VFS removes the
restriction that src and dst files must be on the same superblock.
NFS's copy_file_range() determines if the copy is "intra" or "inter"
and for "inter" it sends the COPY_NOTIFY to the source server. Then,
it would send of an asynchronous COPY to the destination server. If
an application cancels an in-flight COPY, OFFLOAD_CANCEL is sent to
both of the servers.
This patch series also include necessary client-side additions that
are performed by the destination server. The server needs an NFS
open that represents a source file without doing an actual open.
Two function nfs42_ssc_open/nfs42_ssc_close() are introduced to
accomplish it that make use of the VFS's alloc_file_pseudo() to
represent an open.
Also this particular open is marked (NFS_SVC_SSC_COPY_STATE) so
that if the destination server ever to receive stateid errors on
this stateid, it knows not to initiate state recovery (in case
when source server reboots). The recovery must be done by the
client and a new copy must be initiated. Therefore, in this case
the recovery needs to fail with EIO.
v2:
-- VFS changes are to remove VFS check for same copy_file_range
functions before calling it. Instead add the appropriate checks
to the existing implementors of the copy_file_range.
-- Changed the check for the "intra" vs "inter" based on the
serverowner instead of the IPs and merged it with the 1st caller
patch
-- Addressed Anna's comments to ifdef the nfs42_copy_notify in
nfs42.h to make for proper compiling
-- Removed a leftover dprintk from the nfs42_ssc_open
Olga Kornievskaia (13):
VFS permit cross device vfs_copy_file_range
CIFS: add cross-device check for copy_file_range
OverlayFS: add cross-device check for copy_file_range
NFS: add cross file system check for copy_file_range
VFS: Don't copy beyond the end of the file
NFS NFSD defining nl4_servers structure needed by both
NFS add COPY_NOTIFY operation
NFS add ca_source_server<> to COPY
NFS also send OFFLOAD_CANCEL to source server
NFS inter ssc open
NFS skip recovery of copy open on dest server
NFS for "inter" copy treat ESTALE as ENOTSUPP
NFS COPY handle ERR_OFFLOAD_DENIED
Documentation/filesystems/porting | 7 ++
Documentation/filesystems/vfs.txt | 6 +-
fs/cifs/cifsfs.c | 2 +
fs/nfs/nfs42.h | 15 ++-
fs/nfs/nfs42proc.c | 129 ++++++++++++++++++++++---
fs/nfs/nfs42xdr.c | 193 +++++++++++++++++++++++++++++++++++++-
fs/nfs/nfs4_fs.h | 10 ++
fs/nfs/nfs4client.c | 2 +-
fs/nfs/nfs4file.c | 122 +++++++++++++++++++++++-
fs/nfs/nfs4proc.c | 6 +-
fs/nfs/nfs4state.c | 14 +++
fs/nfs/nfs4xdr.c | 1 +
fs/overlayfs/file.c | 3 +
fs/read_write.c | 12 +--
include/linux/nfs4.h | 25 +++++
include/linux/nfs_fs_sb.h | 1 +
include/linux/nfs_xdr.h | 17 ++++
17 files changed, 539 insertions(+), 26 deletions(-)
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
A proposed VFS change removes the check for the files to reside
under the same file system. Instead, a file system driver implementation
is allowed to perform a cross-device copy_file_range() and if
the file system fails to support it instead fallback to doing
a do_splice copy. Therefore, EXDEV is no longer a possible error.
Signed-off-by: Olga Kornievskaia <[email protected]>
---
man2/copy_file_range.2 | 7 ++-----
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/man2/copy_file_range.2 b/man2/copy_file_range.2
index 20374ab..723b2d0 100644
--- a/man2/copy_file_range.2
+++ b/man2/copy_file_range.2
@@ -39,7 +39,8 @@ The
.BR copy_file_range ()
system call performs an in-kernel copy between two file descriptors
without the additional cost of transferring data from the kernel to user space
-and then back into the kernel.
+and then back into the kernel. Since kernel version 4.21(???) passed in
+file descriptors are not required to be under the same mounted file system.
It copies up to
.I len
bytes of data from file descriptor
@@ -128,10 +129,6 @@ Out of memory.
.B ENOSPC
There is not enough space on the target filesystem to complete the copy.
.TP
-.B EXDEV
-The files referred to by
-.IR file_in " and " file_out
-are not on the same mounted filesystem.
.SH VERSIONS
The
.BR copy_file_range ()
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
VFS copy_file_range was relaxed to allow for cross-device copy.
Add a check for now, to disallow cross-device copy offload until
appropriate support is added.
Signed-off-by: Olga Kornievskaia <[email protected]>
---
fs/cifs/cifsfs.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c
index 7065426..1f41e74 100644
--- a/fs/cifs/cifsfs.c
+++ b/fs/cifs/cifsfs.c
@@ -1114,6 +1114,8 @@ static ssize_t cifs_copy_file_range(struct file *src_file, loff_t off,
unsigned int xid = get_xid();
ssize_t rc;
+ if (src_file->f_inode->i_sb != dst_file->f_inode->i_sb)
+ return -EXDEV;
rc = cifs_file_copychunk_range(xid, src_file, off, dst_file, destoff,
len, flags);
free_xid(xid);
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
VFS copy_file_range was relaxed to allow for cross-device copy.
Add a check for now, to disallow cross-device copy offload until
appropriate support is added.
Signed-off-by: Olga Kornievskaia <[email protected]>
`
---
fs/overlayfs/file.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/fs/overlayfs/file.c b/fs/overlayfs/file.c
index aeaefd2..5282853 100644
--- a/fs/overlayfs/file.c
+++ b/fs/overlayfs/file.c
@@ -483,6 +483,9 @@ static ssize_t ovl_copy_file_range(struct file *file_in, loff_t pos_in,
struct file *file_out, loff_t pos_out,
size_t len, unsigned int flags)
{
+ if (file_in->f_inode->i_sb != file_out->f_inode->i_sb)
+ return -EXDEV;
+
return ovl_copyfile(file_in, pos_in, file_out, pos_out, len, flags,
OVL_COPY);
}
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
These structures are needed by COPY_NOTIFY on the client and needed
by the nfsd as well
Signed-off-by: Olga Kornievskaia <[email protected]>
---
include/linux/nfs4.h | 24 ++++++++++++++++++++++++
1 file changed, 24 insertions(+)
diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h
index 1b06f0b..4d76f87 100644
--- a/include/linux/nfs4.h
+++ b/include/linux/nfs4.h
@@ -16,6 +16,7 @@
#include <linux/list.h>
#include <linux/uidgid.h>
#include <uapi/linux/nfs4.h>
+#include <linux/sunrpc/msg_prot.h>
enum nfs4_acl_whotype {
NFS4_ACL_WHO_NAMED = 0,
@@ -672,4 +673,27 @@ struct nfs4_op_map {
} u;
};
+struct nfs42_netaddr {
+ char netid[RPCBIND_MAXNETIDLEN];
+ char addr[RPCBIND_MAXUADDRLEN + 1];
+ u32 netid_len;
+ u32 addr_len;
+};
+
+enum netloc_type4 {
+ NL4_NAME = 1,
+ NL4_URL = 2,
+ NL4_NETADDR = 3,
+};
+
+struct nl4_server {
+ enum netloc_type4 nl4_type;
+ union {
+ struct { /* NL4_NAME, NL4_URL */
+ int nl4_str_sz;
+ char nl4_str[NFS4_OPAQUE_LIMIT + 1];
+ };
+ struct nfs42_netaddr nl4_addr; /* NL4_NETADDR */
+ } u;
+};
#endif
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
VFS copy_file_range was relaxed to allow for cross-device copy.
Add a check to disallow cross file systems copy offload, both
files are expected to be of NFS type.
Signed-off-by: Olga Kornievskaia <[email protected]>
---
fs/nfs/nfs4file.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c
index 4288a6e..7137e7b 100644
--- a/fs/nfs/nfs4file.c
+++ b/fs/nfs/nfs4file.c
@@ -135,6 +135,9 @@ static ssize_t nfs4_copy_file_range(struct file *file_in, loff_t pos_in,
{
ssize_t ret;
+ if (file_in->f_inode->i_sb->s_type != file_out->f_inode->i_sb->s_type)
+ return -EXDEV;
+
if (file_inode(file_in) == file_inode(file_out))
return -EINVAL;
retry:
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
Enforce the check that source offset can't be beyond the end of the file.
Signed-off-by: Anna Schumaker <[email protected]>
---
fs/read_write.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/fs/read_write.c b/fs/read_write.c
index fb4ffca..7bdbf39 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -1570,6 +1570,9 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
if (unlikely(ret))
return ret;
+ if (pos_in >= i_size_read(inode_in))
+ return -EINVAL;
+
if (!(file_in->f_mode & FMODE_READ) ||
!(file_out->f_mode & FMODE_WRITE) ||
(file_out->f_flags & O_APPEND))
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
Try using the delegation stateid, then the open stateid.
Only NL4_NETATTR, No support for NL4_NAME and NL4_URL.
Allow only one source server address to be returned for now.
To distinguish between same server copy offload ("intra") and
a copy between different server ("inter"), do a check of server
owner identity.
Signed-off-by: Andy Adamson <[email protected]>
Signed-off-by: Olga Kornievskaia <[email protected]>
---
fs/nfs/nfs42.h | 12 +++
fs/nfs/nfs42proc.c | 91 +++++++++++++++++++++++
fs/nfs/nfs42xdr.c | 181 ++++++++++++++++++++++++++++++++++++++++++++++
fs/nfs/nfs4_fs.h | 2 +
fs/nfs/nfs4client.c | 2 +-
fs/nfs/nfs4file.c | 14 ++++
fs/nfs/nfs4proc.c | 1 +
fs/nfs/nfs4xdr.c | 1 +
include/linux/nfs4.h | 1 +
include/linux/nfs_fs_sb.h | 1 +
include/linux/nfs_xdr.h | 16 ++++
11 files changed, 321 insertions(+), 1 deletion(-)
diff --git a/fs/nfs/nfs42.h b/fs/nfs/nfs42.h
index 19ec38f8..bbe49a3 100644
--- a/fs/nfs/nfs42.h
+++ b/fs/nfs/nfs42.h
@@ -13,6 +13,7 @@
#define PNFS_LAYOUTSTATS_MAXDEV (4)
/* nfs4.2proc.c */
+#ifdef CONFIG_NFS_V4_2
int nfs42_proc_allocate(struct file *, loff_t, loff_t);
ssize_t nfs42_proc_copy(struct file *, loff_t, struct file *, loff_t, size_t);
int nfs42_proc_deallocate(struct file *, loff_t, loff_t);
@@ -20,5 +21,16 @@
int nfs42_proc_layoutstats_generic(struct nfs_server *,
struct nfs42_layoutstat_data *);
int nfs42_proc_clone(struct file *, struct file *, loff_t, loff_t, loff_t);
+int nfs42_proc_copy_notify(struct file *, struct file *,
+ struct nfs42_copy_notify_res *);
+static inline bool nfs42_files_from_same_server(struct file *in,
+ struct file *out)
+{
+ struct nfs_client *c_in = (NFS_SERVER(file_inode(in)))->nfs_client;
+ struct nfs_client *c_out = (NFS_SERVER(file_inode(out)))->nfs_client;
+ return nfs4_check_serverowner_major_id(c_in->cl_serverowner,
+ c_out->cl_serverowner);
+}
+#endif /* CONFIG_NFS_V4_2 */
#endif /* __LINUX_FS_NFS_NFS4_2_H */
diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c
index ac5b784..b1c57a4 100644
--- a/fs/nfs/nfs42proc.c
+++ b/fs/nfs/nfs42proc.c
@@ -3,6 +3,7 @@
* Copyright (c) 2014 Anna Schumaker <[email protected]>
*/
#include <linux/fs.h>
+#include <linux/sunrpc/addr.h>
#include <linux/sunrpc/sched.h>
#include <linux/nfs.h>
#include <linux/nfs3.h>
@@ -15,10 +16,30 @@
#include "pnfs.h"
#include "nfs4session.h"
#include "internal.h"
+#include "delegation.h"
#define NFSDBG_FACILITY NFSDBG_PROC
static int nfs42_do_offload_cancel_async(struct file *dst, nfs4_stateid *std);
+static void nfs42_set_netaddr(struct file *filep, struct nfs42_netaddr *naddr)
+{
+ struct nfs_client *clp = (NFS_SERVER(file_inode(filep)))->nfs_client;
+ unsigned short port = 2049;
+
+ rcu_read_lock();
+ naddr->netid_len = scnprintf(naddr->netid,
+ sizeof(naddr->netid), "%s",
+ rpc_peeraddr2str(clp->cl_rpcclient,
+ RPC_DISPLAY_NETID));
+ naddr->addr_len = scnprintf(naddr->addr,
+ sizeof(naddr->addr),
+ "%s.%u.%u",
+ rpc_peeraddr2str(clp->cl_rpcclient,
+ RPC_DISPLAY_ADDR),
+ port >> 8, port & 255);
+ rcu_read_unlock();
+}
+
static int _nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep,
struct nfs_lock_context *lock, loff_t offset, loff_t len)
{
@@ -461,6 +482,76 @@ static int nfs42_do_offload_cancel_async(struct file *dst,
return status;
}
+int _nfs42_proc_copy_notify(struct file *src, struct file *dst,
+ struct nfs42_copy_notify_args *args,
+ struct nfs42_copy_notify_res *res)
+{
+ struct nfs_server *src_server = NFS_SERVER(file_inode(src));
+ struct rpc_message msg = {
+ .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY_NOTIFY],
+ .rpc_argp = args,
+ .rpc_resp = res,
+ };
+ int status;
+ struct nfs_open_context *ctx;
+ struct nfs_lock_context *l_ctx;
+
+ ctx = get_nfs_open_context(nfs_file_open_context(src));
+ l_ctx = nfs_get_lock_context(ctx);
+ if (IS_ERR(l_ctx))
+ return PTR_ERR(l_ctx);
+
+ status = nfs4_set_rw_stateid(&args->cna_src_stateid, ctx, l_ctx,
+ FMODE_READ);
+ nfs_put_lock_context(l_ctx);
+ if (status)
+ return status;
+
+ status = nfs4_call_sync(src_server->client, src_server, &msg,
+ &args->cna_seq_args, &res->cnr_seq_res, 0);
+ if (status == -ENOTSUPP)
+ src_server->caps &= ~NFS_CAP_COPY_NOTIFY;
+
+ put_nfs_open_context(nfs_file_open_context(src));
+ return status;
+}
+
+int nfs42_proc_copy_notify(struct file *src, struct file *dst,
+ struct nfs42_copy_notify_res *res)
+{
+ struct nfs_server *src_server = NFS_SERVER(file_inode(src));
+ struct nfs42_copy_notify_args *args;
+ struct nfs4_exception exception = {
+ .inode = file_inode(src),
+ };
+ int status;
+
+ if (!(src_server->caps & NFS_CAP_COPY_NOTIFY))
+ return -EOPNOTSUPP;
+
+ args = kzalloc(sizeof(struct nfs42_copy_notify_args), GFP_NOFS);
+ if (args == NULL)
+ return -ENOMEM;
+
+ args->cna_src_fh = NFS_FH(file_inode(src)),
+ args->cna_dst.nl4_type = NL4_NETADDR;
+ nfs42_set_netaddr(dst, &args->cna_dst.u.nl4_addr);
+ exception.stateid = &args->cna_src_stateid;
+
+ do {
+ status = _nfs42_proc_copy_notify(src, dst, args, res);
+ if (status == -ENOTSUPP) {
+ status = -EOPNOTSUPP;
+ goto out;
+ }
+ status = nfs4_handle_exception(src_server, status, &exception);
+ } while (exception.retry);
+
+out:
+ kfree(args);
+ return status;
+}
+
static loff_t _nfs42_proc_llseek(struct file *filep,
struct nfs_lock_context *lock, loff_t offset, int whence)
{
diff --git a/fs/nfs/nfs42xdr.c b/fs/nfs/nfs42xdr.c
index 69f72ed..e6e7cbf 100644
--- a/fs/nfs/nfs42xdr.c
+++ b/fs/nfs/nfs42xdr.c
@@ -29,6 +29,16 @@
#define encode_offload_cancel_maxsz (op_encode_hdr_maxsz + \
XDR_QUADLEN(NFS4_STATEID_SIZE))
#define decode_offload_cancel_maxsz (op_decode_hdr_maxsz)
+#define encode_copy_notify_maxsz (op_encode_hdr_maxsz + \
+ XDR_QUADLEN(NFS4_STATEID_SIZE) + \
+ 1 + /* nl4_type */ \
+ 1 + XDR_QUADLEN(NFS4_OPAQUE_LIMIT))
+#define decode_copy_notify_maxsz (op_decode_hdr_maxsz + \
+ 3 + /* cnr_lease_time */\
+ XDR_QUADLEN(NFS4_STATEID_SIZE) + \
+ 1 + /* Support 1 cnr_source_server */\
+ 1 + /* nl4_type */ \
+ 1 + XDR_QUADLEN(NFS4_OPAQUE_LIMIT))
#define encode_deallocate_maxsz (op_encode_hdr_maxsz + \
encode_fallocate_maxsz)
#define decode_deallocate_maxsz (op_decode_hdr_maxsz)
@@ -84,6 +94,12 @@
#define NFS4_dec_offload_cancel_sz (compound_decode_hdr_maxsz + \
decode_putfh_maxsz + \
decode_offload_cancel_maxsz)
+#define NFS4_enc_copy_notify_sz (compound_encode_hdr_maxsz + \
+ encode_putfh_maxsz + \
+ encode_copy_notify_maxsz)
+#define NFS4_dec_copy_notify_sz (compound_decode_hdr_maxsz + \
+ decode_putfh_maxsz + \
+ decode_copy_notify_maxsz)
#define NFS4_enc_deallocate_sz (compound_encode_hdr_maxsz + \
encode_putfh_maxsz + \
encode_deallocate_maxsz + \
@@ -137,6 +153,25 @@ static void encode_allocate(struct xdr_stream *xdr,
encode_fallocate(xdr, args);
}
+static void encode_nl4_server(struct xdr_stream *xdr, const struct nl4_server *ns)
+{
+ encode_uint32(xdr, ns->nl4_type);
+ switch (ns->nl4_type) {
+ case NL4_NAME:
+ case NL4_URL:
+ encode_string(xdr, ns->u.nl4_str_sz, ns->u.nl4_str);
+ break;
+ case NL4_NETADDR:
+ encode_string(xdr, ns->u.nl4_addr.netid_len,
+ ns->u.nl4_addr.netid);
+ encode_string(xdr, ns->u.nl4_addr.addr_len,
+ ns->u.nl4_addr.addr);
+ break;
+ default:
+ WARN_ON_ONCE(1);
+ }
+}
+
static void encode_copy(struct xdr_stream *xdr,
const struct nfs42_copy_args *args,
struct compound_hdr *hdr)
@@ -162,6 +197,15 @@ static void encode_offload_cancel(struct xdr_stream *xdr,
encode_nfs4_stateid(xdr, &args->osa_stateid);
}
+static void encode_copy_notify(struct xdr_stream *xdr,
+ const struct nfs42_copy_notify_args *args,
+ struct compound_hdr *hdr)
+{
+ encode_op_hdr(xdr, OP_COPY_NOTIFY, decode_copy_notify_maxsz, hdr);
+ encode_nfs4_stateid(xdr, &args->cna_src_stateid);
+ encode_nl4_server(xdr, &args->cna_dst);
+}
+
static void encode_deallocate(struct xdr_stream *xdr,
const struct nfs42_falloc_args *args,
struct compound_hdr *hdr)
@@ -298,6 +342,25 @@ static void nfs4_xdr_enc_offload_cancel(struct rpc_rqst *req,
}
/*
+ * Encode COPY_NOTIFY request
+ */
+static void nfs4_xdr_enc_copy_notify(struct rpc_rqst *req,
+ struct xdr_stream *xdr,
+ const void *data)
+{
+ const struct nfs42_copy_notify_args *args = data;
+ struct compound_hdr hdr = {
+ .minorversion = nfs4_xdr_minorversion(&args->cna_seq_args),
+ };
+
+ encode_compound_hdr(xdr, req, &hdr);
+ encode_sequence(xdr, &args->cna_seq_args, &hdr);
+ encode_putfh(xdr, args->cna_src_fh, &hdr);
+ encode_copy_notify(xdr, args, &hdr);
+ encode_nops(&hdr);
+}
+
+/*
* Encode DEALLOCATE request
*/
static void nfs4_xdr_enc_deallocate(struct rpc_rqst *req,
@@ -416,6 +479,58 @@ static int decode_write_response(struct xdr_stream *xdr,
return -EIO;
}
+static int decode_nl4_server(struct xdr_stream *xdr, struct nl4_server *ns)
+{
+ struct nfs42_netaddr *naddr;
+ uint32_t dummy;
+ char *dummy_str;
+ __be32 *p;
+ int status;
+
+ /* nl_type */
+ p = xdr_inline_decode(xdr, 4);
+ if (unlikely(!p))
+ return -EIO;
+ ns->nl4_type = be32_to_cpup(p);
+ switch (ns->nl4_type) {
+ case NL4_NAME:
+ case NL4_URL:
+ status = decode_opaque_inline(xdr, &dummy, &dummy_str);
+ if (unlikely(status))
+ return status;
+ if (unlikely(dummy > NFS4_OPAQUE_LIMIT))
+ return -EIO;
+ memcpy(&ns->u.nl4_str, dummy_str, dummy);
+ ns->u.nl4_str_sz = dummy;
+ break;
+ case NL4_NETADDR:
+ naddr = &ns->u.nl4_addr;
+
+ /* netid string */
+ status = decode_opaque_inline(xdr, &dummy, &dummy_str);
+ if (unlikely(status))
+ return status;
+ if (unlikely(dummy > RPCBIND_MAXNETIDLEN))
+ return -EIO;
+ naddr->netid_len = dummy;
+ memcpy(naddr->netid, dummy_str, naddr->netid_len);
+
+ /* uaddr string */
+ status = decode_opaque_inline(xdr, &dummy, &dummy_str);
+ if (unlikely(status))
+ return status;
+ if (unlikely(dummy > RPCBIND_MAXUADDRLEN))
+ return -EIO;
+ naddr->addr_len = dummy;
+ memcpy(naddr->addr, dummy_str, naddr->addr_len);
+ break;
+ default:
+ WARN_ON_ONCE(1);
+ return -EIO;
+ }
+ return 0;
+}
+
static int decode_copy_requirements(struct xdr_stream *xdr,
struct nfs42_copy_res *res) {
__be32 *p;
@@ -458,6 +573,46 @@ static int decode_offload_cancel(struct xdr_stream *xdr,
return decode_op_hdr(xdr, OP_OFFLOAD_CANCEL);
}
+static int decode_copy_notify(struct xdr_stream *xdr,
+ struct nfs42_copy_notify_res *res)
+{
+ __be32 *p;
+ int status, count;
+
+ status = decode_op_hdr(xdr, OP_COPY_NOTIFY);
+ if (status)
+ return status;
+ /* cnr_lease_time */
+ p = xdr_inline_decode(xdr, 12);
+ if (unlikely(!p))
+ goto out_overflow;
+ p = xdr_decode_hyper(p, &res->cnr_lease_time.seconds);
+ res->cnr_lease_time.nseconds = be32_to_cpup(p);
+
+ status = decode_opaque_fixed(xdr, &res->cnr_stateid, NFS4_STATEID_SIZE);
+ if (unlikely(status))
+ goto out_overflow;
+
+ /* number of source addresses */
+ p = xdr_inline_decode(xdr, 4);
+ if (unlikely(!p))
+ goto out_overflow;
+
+ count = be32_to_cpup(p);
+ if (count > 1)
+ pr_warn("NFS: %s: nsvr %d > Supported. Use first servers\n",
+ __func__, count);
+
+ status = decode_nl4_server(xdr, &res->cnr_src);
+ if (unlikely(status))
+ goto out_overflow;
+ return 0;
+
+out_overflow:
+ print_overflow_msg(__func__, xdr);
+ return -EIO;
+}
+
static int decode_deallocate(struct xdr_stream *xdr, struct nfs42_falloc_res *res)
{
return decode_op_hdr(xdr, OP_DEALLOCATE);
@@ -585,6 +740,32 @@ static int nfs4_xdr_dec_offload_cancel(struct rpc_rqst *rqstp,
}
/*
+ * Decode COPY_NOTIFY response
+ */
+static int nfs4_xdr_dec_copy_notify(struct rpc_rqst *rqstp,
+ struct xdr_stream *xdr,
+ void *data)
+{
+ struct nfs42_copy_notify_res *res = data;
+ struct compound_hdr hdr;
+ int status;
+
+ status = decode_compound_hdr(xdr, &hdr);
+ if (status)
+ goto out;
+ status = decode_sequence(xdr, &res->cnr_seq_res, rqstp);
+ if (status)
+ goto out;
+ status = decode_putfh(xdr);
+ if (status)
+ goto out;
+ status = decode_copy_notify(xdr, res);
+
+out:
+ return status;
+}
+
+/*
* Decode DEALLOCATE request
*/
static int nfs4_xdr_dec_deallocate(struct rpc_rqst *rqstp,
diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
index 8d59c96..7d17b31 100644
--- a/fs/nfs/nfs4_fs.h
+++ b/fs/nfs/nfs4_fs.h
@@ -460,6 +460,8 @@ int nfs41_discover_server_trunking(struct nfs_client *clp,
struct nfs_client **, struct rpc_cred *);
extern void nfs4_schedule_session_recovery(struct nfs4_session *, int);
extern void nfs41_notify_server(struct nfs_client *);
+bool nfs4_check_serverowner_major_id(struct nfs41_server_owner *o1,
+ struct nfs41_server_owner *o2);
#else
static inline void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
{
diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c
index 8f53455..ac00eb8 100644
--- a/fs/nfs/nfs4client.c
+++ b/fs/nfs/nfs4client.c
@@ -625,7 +625,7 @@ int nfs40_walk_client_list(struct nfs_client *new,
/*
* Returns true if the server major ids match
*/
-static bool
+bool
nfs4_check_serverowner_major_id(struct nfs41_server_owner *o1,
struct nfs41_server_owner *o2)
{
diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c
index 7137e7b..ae51e38 100644
--- a/fs/nfs/nfs4file.c
+++ b/fs/nfs/nfs4file.c
@@ -133,6 +133,7 @@ static ssize_t nfs4_copy_file_range(struct file *file_in, loff_t pos_in,
struct file *file_out, loff_t pos_out,
size_t count, unsigned int flags)
{
+ struct nfs42_copy_notify_res *cn_resp = NULL;
ssize_t ret;
if (file_in->f_inode->i_sb->s_type != file_out->f_inode->i_sb->s_type)
@@ -141,7 +142,20 @@ static ssize_t nfs4_copy_file_range(struct file *file_in, loff_t pos_in,
if (file_inode(file_in) == file_inode(file_out))
return -EINVAL;
retry:
+ if (!nfs42_files_from_same_server(file_in, file_out)) {
+ cn_resp = kzalloc(sizeof(struct nfs42_copy_notify_res),
+ GFP_NOFS);
+ if (unlikely(cn_resp == NULL))
+ return -ENOMEM;
+
+ ret = nfs42_proc_copy_notify(file_in, file_out, cn_resp);
+ if (ret)
+ goto out;
+ }
+
ret = nfs42_proc_copy(file_in, pos_in, file_out, pos_out, count);
+out:
+ kfree(cn_resp);
if (ret == -EAGAIN)
goto retry;
return ret;
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index db84b4a..fec6e6b 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -9692,6 +9692,7 @@ static bool nfs4_match_stateid(const nfs4_stateid *s1,
| NFS_CAP_ALLOCATE
| NFS_CAP_COPY
| NFS_CAP_OFFLOAD_CANCEL
+ | NFS_CAP_COPY_NOTIFY
| NFS_CAP_DEALLOCATE
| NFS_CAP_SEEK
| NFS_CAP_LAYOUTSTATS
diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
index b7bde12..2163900 100644
--- a/fs/nfs/nfs4xdr.c
+++ b/fs/nfs/nfs4xdr.c
@@ -7790,6 +7790,7 @@ int nfs4_decode_dirent(struct xdr_stream *xdr, struct nfs_entry *entry,
PROC42(CLONE, enc_clone, dec_clone),
PROC42(COPY, enc_copy, dec_copy),
PROC42(OFFLOAD_CANCEL, enc_offload_cancel, dec_offload_cancel),
+ PROC42(COPY_NOTIFY, enc_copy_notify, dec_copy_notify),
PROC(LOOKUPP, enc_lookupp, dec_lookupp),
};
diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h
index 4d76f87..d80b25e 100644
--- a/include/linux/nfs4.h
+++ b/include/linux/nfs4.h
@@ -537,6 +537,7 @@ enum {
NFSPROC4_CLNT_CLONE,
NFSPROC4_CLNT_COPY,
NFSPROC4_CLNT_OFFLOAD_CANCEL,
+ NFSPROC4_CLNT_COPY_NOTIFY,
NFSPROC4_CLNT_LOOKUPP,
};
diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h
index 0fc0b91..e5d89ff 100644
--- a/include/linux/nfs_fs_sb.h
+++ b/include/linux/nfs_fs_sb.h
@@ -261,5 +261,6 @@ struct nfs_server {
#define NFS_CAP_CLONE (1U << 23)
#define NFS_CAP_COPY (1U << 24)
#define NFS_CAP_OFFLOAD_CANCEL (1U << 25)
+#define NFS_CAP_COPY_NOTIFY (1U << 26)
#endif
diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h
index 0e01625..dfc59bc 100644
--- a/include/linux/nfs_xdr.h
+++ b/include/linux/nfs_xdr.h
@@ -1428,6 +1428,22 @@ struct nfs42_offload_status_res {
int osr_status;
};
+struct nfs42_copy_notify_args {
+ struct nfs4_sequence_args cna_seq_args;
+
+ struct nfs_fh *cna_src_fh;
+ nfs4_stateid cna_src_stateid;
+ struct nl4_server cna_dst;
+};
+
+struct nfs42_copy_notify_res {
+ struct nfs4_sequence_res cnr_seq_res;
+
+ struct nfstime4 cnr_lease_time;
+ nfs4_stateid cnr_stateid;
+ struct nl4_server cnr_src;
+};
+
struct nfs42_seek_args {
struct nfs4_sequence_args seq_args;
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
Support only one source server address: the same address that
the client and source server use.
Signed-off-by: Andy Adamson <[email protected]>
Signed-off-by: Olga Kornievskaia <[email protected]>
---
fs/nfs/nfs42.h | 3 ++-
fs/nfs/nfs42proc.c | 26 +++++++++++++++++---------
fs/nfs/nfs42xdr.c | 12 ++++++++++--
fs/nfs/nfs4file.c | 7 ++++++-
include/linux/nfs_xdr.h | 1 +
5 files changed, 36 insertions(+), 13 deletions(-)
diff --git a/fs/nfs/nfs42.h b/fs/nfs/nfs42.h
index bbe49a3..28dcee5 100644
--- a/fs/nfs/nfs42.h
+++ b/fs/nfs/nfs42.h
@@ -15,7 +15,8 @@
/* nfs4.2proc.c */
#ifdef CONFIG_NFS_V4_2
int nfs42_proc_allocate(struct file *, loff_t, loff_t);
-ssize_t nfs42_proc_copy(struct file *, loff_t, struct file *, loff_t, size_t);
+ssize_t nfs42_proc_copy(struct file *, loff_t, struct file *, loff_t, size_t,
+ struct nl4_server *, nfs4_stateid *);
int nfs42_proc_deallocate(struct file *, loff_t, loff_t);
loff_t nfs42_proc_llseek(struct file *, loff_t, int);
int nfs42_proc_layoutstats_generic(struct nfs_server *,
diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c
index b1c57a4..bb9e799 100644
--- a/fs/nfs/nfs42proc.c
+++ b/fs/nfs/nfs42proc.c
@@ -242,7 +242,9 @@ static ssize_t _nfs42_proc_copy(struct file *src,
struct file *dst,
struct nfs_lock_context *dst_lock,
struct nfs42_copy_args *args,
- struct nfs42_copy_res *res)
+ struct nfs42_copy_res *res,
+ struct nl4_server *nss,
+ nfs4_stateid *cnr_stateid)
{
struct rpc_message msg = {
.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY],
@@ -256,11 +258,15 @@ static ssize_t _nfs42_proc_copy(struct file *src,
size_t count = args->count;
ssize_t status;
- status = nfs4_set_rw_stateid(&args->src_stateid, src_lock->open_context,
- src_lock, FMODE_READ);
- if (status)
- return status;
-
+ if (nss) {
+ args->cp_src = nss;
+ nfs4_stateid_copy(&args->src_stateid, cnr_stateid);
+ } else {
+ status = nfs4_set_rw_stateid(&args->src_stateid,
+ src_lock->open_context, src_lock, FMODE_READ);
+ if (status)
+ return status;
+ }
status = nfs_filemap_write_and_wait_range(file_inode(src)->i_mapping,
pos_src, pos_src + (loff_t)count - 1);
if (status)
@@ -324,8 +330,9 @@ static ssize_t _nfs42_proc_copy(struct file *src,
}
ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src,
- struct file *dst, loff_t pos_dst,
- size_t count)
+ struct file *dst, loff_t pos_dst, size_t count,
+ struct nl4_server *nss,
+ nfs4_stateid *cnr_stateid)
{
struct nfs_server *server = NFS_SERVER(file_inode(dst));
struct nfs_lock_context *src_lock;
@@ -370,7 +377,8 @@ ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src,
inode_lock(file_inode(dst));
err = _nfs42_proc_copy(src, src_lock,
dst, dst_lock,
- &args, &res);
+ &args, &res,
+ nss, cnr_stateid);
inode_unlock(file_inode(dst));
if (err >= 0)
diff --git a/fs/nfs/nfs42xdr.c b/fs/nfs/nfs42xdr.c
index e6e7cbf..c96c3f8 100644
--- a/fs/nfs/nfs42xdr.c
+++ b/fs/nfs/nfs42xdr.c
@@ -21,7 +21,10 @@
#define encode_copy_maxsz (op_encode_hdr_maxsz + \
XDR_QUADLEN(NFS4_STATEID_SIZE) + \
XDR_QUADLEN(NFS4_STATEID_SIZE) + \
- 2 + 2 + 2 + 1 + 1 + 1)
+ 2 + 2 + 2 + 1 + 1 + 1 +\
+ 1 + /* One cnr_source_server */\
+ 1 + /* nl4_type */ \
+ 1 + XDR_QUADLEN(NFS4_OPAQUE_LIMIT))
#define decode_copy_maxsz (op_decode_hdr_maxsz + \
NFS42_WRITE_RES_SIZE + \
1 /* cr_consecutive */ + \
@@ -186,7 +189,12 @@ static void encode_copy(struct xdr_stream *xdr,
encode_uint32(xdr, 1); /* consecutive = true */
encode_uint32(xdr, args->sync);
- encode_uint32(xdr, 0); /* src server list */
+ if (args->cp_src == NULL) { /* intra-ssc */
+ encode_uint32(xdr, 0); /* no src server list */
+ return;
+ }
+ encode_uint32(xdr, 1); /* supporting 1 server */
+ encode_nl4_server(xdr, args->cp_src);
}
static void encode_offload_cancel(struct xdr_stream *xdr,
diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c
index ae51e38..703b56f 100644
--- a/fs/nfs/nfs4file.c
+++ b/fs/nfs/nfs4file.c
@@ -134,6 +134,8 @@ static ssize_t nfs4_copy_file_range(struct file *file_in, loff_t pos_in,
size_t count, unsigned int flags)
{
struct nfs42_copy_notify_res *cn_resp = NULL;
+ struct nl4_server *nss = NULL;
+ nfs4_stateid *cnrs = NULL;
ssize_t ret;
if (file_in->f_inode->i_sb->s_type != file_out->f_inode->i_sb->s_type)
@@ -151,9 +153,12 @@ static ssize_t nfs4_copy_file_range(struct file *file_in, loff_t pos_in,
ret = nfs42_proc_copy_notify(file_in, file_out, cn_resp);
if (ret)
goto out;
+ nss = &cn_resp->cnr_src;
+ cnrs = &cn_resp->cnr_stateid;
}
- ret = nfs42_proc_copy(file_in, pos_in, file_out, pos_out, count);
+ ret = nfs42_proc_copy(file_in, pos_in, file_out, pos_out, count, nss,
+ cnrs);
out:
kfree(cn_resp);
if (ret == -EAGAIN)
diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h
index dfc59bc..3a40b17 100644
--- a/include/linux/nfs_xdr.h
+++ b/include/linux/nfs_xdr.h
@@ -1400,6 +1400,7 @@ struct nfs42_copy_args {
u64 count;
bool sync;
+ struct nl4_server *cp_src;
};
struct nfs42_write_res {
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
In case of copy is cancelled, also send OFFLOAD_CANCEL to the source
server.
Signed-off-by: Olga Kornievskaia <[email protected]>
---
fs/nfs/nfs42proc.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c
index bb9e799..98fe00b 100644
--- a/fs/nfs/nfs42proc.c
+++ b/fs/nfs/nfs42proc.c
@@ -205,12 +205,14 @@ static int handle_async_copy(struct nfs42_copy_res *res,
memcpy(&res->write_res.verifier, ©->verf, sizeof(copy->verf));
status = -copy->error;
+out_free:
kfree(copy);
return status;
out_cancel:
nfs42_do_offload_cancel_async(dst, ©->stateid);
- kfree(copy);
- return status;
+ if (!nfs42_files_from_same_server(src, dst))
+ nfs42_do_offload_cancel_async(src, src_stateid);
+ goto out_free;
}
static int process_copy_commit(struct file *dst, loff_t pos_dst,
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
NFSv4.2 inter server to server copy requires the destination server to
READ the data from the source server using the provided stateid and
file handle.
Given an NFSv4 stateid and filehandle from the COPY operaion, provide the
destination server with an NFS client function to create a struct file
suitable for the destiniation server to READ the data to be copied.
Signed-off-by: Andy Adamson <[email protected]>
Signed-off-by: Olga Kornievskaia <[email protected]>
---
fs/nfs/nfs4_fs.h | 7 ++++
fs/nfs/nfs4file.c | 97 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
fs/nfs/nfs4proc.c | 5 ++-
3 files changed, 106 insertions(+), 3 deletions(-)
diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
index 7d17b31..9c566a4 100644
--- a/fs/nfs/nfs4_fs.h
+++ b/fs/nfs/nfs4_fs.h
@@ -307,6 +307,13 @@ extern int nfs4_set_rw_stateid(nfs4_stateid *stateid,
const struct nfs_open_context *ctx,
const struct nfs_lock_context *l_ctx,
fmode_t fmode);
+extern int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
+ struct nfs_fattr *fattr, struct nfs4_label *label,
+ struct inode *inode);
+extern int update_open_stateid(struct nfs4_state *state,
+ const nfs4_stateid *open_stateid,
+ const nfs4_stateid *deleg_stateid,
+ fmode_t fmode);
#if defined(CONFIG_NFS_V4_1)
extern int nfs41_sequence_done(struct rpc_task *, struct nfs4_sequence_res *);
diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c
index 703b56f..300880f 100644
--- a/fs/nfs/nfs4file.c
+++ b/fs/nfs/nfs4file.c
@@ -8,6 +8,7 @@
#include <linux/file.h>
#include <linux/falloc.h>
#include <linux/nfs_fs.h>
+#include <linux/file.h>
#include "delegation.h"
#include "internal.h"
#include "iostat.h"
@@ -264,6 +265,102 @@ static int nfs42_clone_file_range(struct file *src_file, loff_t src_off,
out:
return ret;
}
+
+static int read_name_gen = 1;
+#define SSC_READ_NAME_BODY "ssc_read_%d"
+
+struct file *
+nfs42_ssc_open(struct vfsmount *ss_mnt, struct nfs_fh *src_fh,
+ nfs4_stateid *stateid)
+{
+ struct nfs_fattr fattr;
+ struct file *filep, *res;
+ struct nfs_server *server;
+ struct inode *r_ino = NULL;
+ struct nfs_open_context *ctx;
+ struct nfs4_state_owner *sp;
+ char *read_name;
+ int len, status = 0;
+
+ server = NFS_SERVER(ss_mnt->mnt_root->d_inode);
+
+ nfs_fattr_init(&fattr);
+
+ status = nfs4_proc_getattr(server, src_fh, &fattr, NULL, NULL);
+ if (status < 0) {
+ res = ERR_PTR(status);
+ goto out;
+ }
+
+ res = ERR_PTR(-ENOMEM);
+ len = strlen(SSC_READ_NAME_BODY) + 16;
+ read_name = kzalloc(len, GFP_NOFS);
+ if (read_name == NULL)
+ goto out;
+ snprintf(read_name, len, SSC_READ_NAME_BODY, read_name_gen++);
+
+ r_ino = nfs_fhget(ss_mnt->mnt_root->d_inode->i_sb, src_fh, &fattr,
+ NULL);
+ if (IS_ERR(r_ino)) {
+ res = ERR_CAST(r_ino);
+ goto out;
+ }
+
+ filep = alloc_file_pseudo(r_ino, ss_mnt, read_name, FMODE_READ,
+ r_ino->i_fop);
+ if (IS_ERR(filep)) {
+ res = ERR_CAST(filep);
+ goto out;
+ }
+ filep->f_mode |= FMODE_READ;
+
+ ctx = alloc_nfs_open_context(filep->f_path.dentry, filep->f_mode,
+ filep);
+ if (IS_ERR(ctx)) {
+ res = ERR_CAST(ctx);
+ goto out_filep;
+ }
+
+ res = ERR_PTR(-EINVAL);
+ sp = nfs4_get_state_owner(server, ctx->cred, GFP_KERNEL);
+ if (sp == NULL)
+ goto out_ctx;
+
+ ctx->state = nfs4_get_open_state(r_ino, sp);
+ if (ctx->state == NULL)
+ goto out_stateowner;
+
+ set_bit(NFS_OPEN_STATE, &ctx->state->flags);
+ memcpy(&ctx->state->open_stateid.other, &stateid->other,
+ NFS4_STATEID_OTHER_SIZE);
+ update_open_stateid(ctx->state, stateid, NULL, filep->f_mode);
+
+ nfs_file_set_open_context(filep, ctx);
+ put_nfs_open_context(ctx);
+
+ file_ra_state_init(&filep->f_ra, filep->f_mapping->host->i_mapping);
+ res = filep;
+out:
+ dprintk("<-- %s error %ld filep %p r_ino %p\n",
+ __func__, IS_ERR(res) ? PTR_ERR(res) : 0, res, r_ino);
+
+ return res;
+out_stateowner:
+ nfs4_put_state_owner(sp);
+out_ctx:
+ put_nfs_open_context(ctx);
+out_filep:
+ fput(filep);
+ goto out;
+}
+EXPORT_SYMBOL_GPL(nfs42_ssc_open);
+void nfs42_ssc_close(struct file *filep)
+{
+ struct nfs_open_context *ctx = nfs_file_open_context(filep);
+
+ ctx->state->flags = 0;
+}
+EXPORT_SYMBOL_GPL(nfs42_ssc_close);
#endif /* CONFIG_NFS_V4_2 */
const struct file_operations nfs4_file_operations = {
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index fec6e6b..e5178b2f 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -91,7 +91,6 @@
static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
-static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label, struct inode *inode);
static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
struct nfs_fattr *fattr, struct iattr *sattr,
@@ -1653,7 +1652,7 @@ static void nfs_state_clear_delegation(struct nfs4_state *state)
write_sequnlock(&state->seqlock);
}
-static int update_open_stateid(struct nfs4_state *state,
+int update_open_stateid(struct nfs4_state *state,
const nfs4_stateid *open_stateid,
const nfs4_stateid *delegation,
fmode_t fmode)
@@ -3936,7 +3935,7 @@ static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
}
-static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
+int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
struct nfs_fattr *fattr, struct nfs4_label *label,
struct inode *inode)
{
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
If the client sends an "inter" copy to the destination server but
it only supports "intra" copy, it can return ESTALE (since it
doesn't know anything about the file handle from the other server
and does not recognize the special case of "inter" copy). Translate
this error as ENOTSUPP and also send OFFLOAD_CANCEL to teh source
server so that it can clean up state.
Signed-off-by: Olga Kornievskaia <[email protected]>
---
fs/nfs/nfs42proc.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c
index 98fe00b..00809b3 100644
--- a/fs/nfs/nfs42proc.c
+++ b/fs/nfs/nfs42proc.c
@@ -395,6 +395,11 @@ ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src,
args.sync = true;
dst_exception.retry = 1;
continue;
+ } else if (err == -ESTALE &&
+ !nfs42_files_from_same_server(src, dst)) {
+ nfs42_do_offload_cancel_async(src, &args.src_stateid);
+ err = -EOPNOTSUPP;
+ break;
}
err2 = nfs4_handle_exception(server, err, &src_exception);
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
If server sends ERR_OFFLOAD_DENIED error, the client must fall
back on doing copy the normal way. Return ENOTSUPP to the vfs and
fallback to regular copy.
Signed-off-by: Olga Kornievskaia <[email protected]>
---
fs/nfs/nfs42proc.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c
index 00809b3..c7c2ffa 100644
--- a/fs/nfs/nfs42proc.c
+++ b/fs/nfs/nfs42proc.c
@@ -395,7 +395,8 @@ ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src,
args.sync = true;
dst_exception.retry = 1;
continue;
- } else if (err == -ESTALE &&
+ } else if ((err == -ESTALE ||
+ err == -NFS4ERR_OFFLOAD_DENIED) &&
!nfs42_files_from_same_server(src, dst)) {
nfs42_do_offload_cancel_async(src, &args.src_stateid);
err = -EOPNOTSUPP;
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
Mark the open created for the source file on the destination
server. Then if this open is going thru a recovery, then fail
the recovery as we don't need to be recoving a "fake" open.
We need to fail the ongoing READs and vfs_copy_file_range().
Signed-off-by: Olga Kornievskaia <[email protected]>
---
fs/nfs/nfs4_fs.h | 1 +
fs/nfs/nfs4file.c | 1 +
fs/nfs/nfs4state.c | 14 ++++++++++++++
3 files changed, 16 insertions(+)
diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
index 9c566a4..482754d 100644
--- a/fs/nfs/nfs4_fs.h
+++ b/fs/nfs/nfs4_fs.h
@@ -165,6 +165,7 @@ enum {
NFS_STATE_CHANGE_WAIT, /* A state changing operation is outstanding */
#ifdef CONFIG_NFS_V4_2
NFS_CLNT_DST_SSC_COPY_STATE, /* dst server open state on client*/
+ NFS_SRV_SSC_COPY_STATE, /* ssc state on the dst server */
#endif /* CONFIG_NFS_V4_2 */
};
diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c
index 300880f..480b384 100644
--- a/fs/nfs/nfs4file.c
+++ b/fs/nfs/nfs4file.c
@@ -330,6 +330,7 @@ struct file *
if (ctx->state == NULL)
goto out_stateowner;
+ set_bit(NFS_SRV_SSC_COPY_STATE, &ctx->state->flags);
set_bit(NFS_OPEN_STATE, &ctx->state->flags);
memcpy(&ctx->state->open_stateid.other, &stateid->other,
NFS4_STATEID_OTHER_SIZE);
diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c
index 62ae0fd..b0b82c6 100644
--- a/fs/nfs/nfs4state.c
+++ b/fs/nfs/nfs4state.c
@@ -1606,6 +1606,9 @@ static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs
{
struct nfs4_state *state;
int status = 0;
+#ifdef CONFIG_NFS_V4_2
+ bool found_ssc_copy_state = false;
+#endif /* CONFIG_NFS_V4_2 */
/* Note: we rely on the sp->so_states list being ordered
* so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
@@ -1625,6 +1628,13 @@ static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs
continue;
if (state->state == 0)
continue;
+#ifdef CONFIG_NFS_V4_2
+ if (test_bit(NFS_SRV_SSC_COPY_STATE, &state->flags)) {
+ nfs4_state_mark_recovery_failed(state, -EIO);
+ found_ssc_copy_state = true;
+ continue;
+ }
+#endif /* CONFIG_NFS_V4_2 */
refcount_inc(&state->count);
spin_unlock(&sp->so_lock);
status = __nfs4_reclaim_open_state(sp, state, ops);
@@ -1671,6 +1681,10 @@ static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs
}
raw_write_seqcount_end(&sp->so_reclaim_seqcount);
spin_unlock(&sp->so_lock);
+#ifdef CONFIG_NFS_V4_2
+ if (found_ssc_copy_state)
+ return -EIO;
+#endif /* CONFIG_NFS_V4_2 */
return 0;
out_err:
nfs4_put_open_state(state);
--
1.8.3.1
From: Olga Kornievskaia <[email protected]>
This patch removes the check for source and destination files to
come from the same superblock. This feature was of interest to
NFS as well as CIFS communities.
Specifically, this feature is needed to allow for NFSv4.2 copy offload
to be done between different NFSv4.2 servers. SMBv3 copy offload between
different servers would be able to use this as well.
Removal of the check implies that passed in source and destination
files can come from different superblocks of the same file system
type or different. It is upto each individual copy_file_range()
file system implementation to decide what type of copy it is
capable of doing and return -EXDEV in cases support is lacking.
There are 3 known implementator of copy_file_range() f_op: NFS,
CIFS, OverlayFS. NFS and CIFS are interested to support cross-device
copy offload but do not support cross file system types copy offload.
Following patches will add appropriate checks in each of the drivers.
If the copy_file_range() errors with EXDEV, the code would fallback
on doing do_splice_direct() copying which in itself is beneficial.
Adding wording to the vfs.txt and porting documentation about the
new support for cross-device copy offload.
Signed-off-by: Olga Kornievskaia <[email protected]>
---
Documentation/filesystems/porting | 7 +++++++
Documentation/filesystems/vfs.txt | 6 +++++-
fs/read_write.c | 9 +++------
3 files changed, 15 insertions(+), 7 deletions(-)
diff --git a/Documentation/filesystems/porting b/Documentation/filesystems/porting
index 7b7b845..ebb4954 100644
--- a/Documentation/filesystems/porting
+++ b/Documentation/filesystems/porting
@@ -622,3 +622,10 @@ in your dentry operations instead.
alloc_file_clone(file, flags, ops) does not affect any caller's references.
On success you get a new struct file sharing the mount/dentry with the
original, on failure - ERR_PTR().
+--
+[mandatory]
+ ->copy_file_range() may now be passed files which belong to two
+ different superblocks of the same file system type or which belong
+ to two different filesystems types all together. As before, the
+ destination's copy_file_range() is the function which is called.
+ If it cannot copy ranges from the source, it should return -EXDEV.
diff --git a/Documentation/filesystems/vfs.txt b/Documentation/filesystems/vfs.txt
index a6c6a8a..34c0e8c 100644
--- a/Documentation/filesystems/vfs.txt
+++ b/Documentation/filesystems/vfs.txt
@@ -1,5 +1,6 @@
Overview of the Linux Virtual File System
+- [fs] nfs: Don't let readdirplus revalidate an inode that was marked as stale (Benjamin Coddington) [1429514 1416532]
Original author: Richard Gooch <[email protected]>
@@ -958,7 +959,10 @@ otherwise noted.
fallocate: called by the VFS to preallocate blocks or punch a hole.
- copy_file_range: called by the copy_file_range(2) system call.
+ copy_file_range: called by copy_file_range(2) system call. This method
+ works on two file descriptors that might reside on
+ different superblocks which might belong to file systems
+ of different types.
clone_file_range: called by the ioctl(2) system call for FICLONERANGE and
FICLONE commands.
diff --git a/fs/read_write.c b/fs/read_write.c
index 39b4a21..fb4ffca 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -1575,10 +1575,6 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
(file_out->f_flags & O_APPEND))
return -EBADF;
- /* this could be relaxed once a method supports cross-fs copies */
- if (inode_in->i_sb != inode_out->i_sb)
- return -EXDEV;
-
if (len == 0)
return 0;
@@ -1588,7 +1584,8 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
* Try cloning first, this is supported by more file systems, and
* more efficient if both clone and copy are supported (e.g. NFS).
*/
- if (file_in->f_op->clone_file_range) {
+ if (inode_in->i_sb == inode_out->i_sb &&
+ file_in->f_op->clone_file_range) {
ret = file_in->f_op->clone_file_range(file_in, pos_in,
file_out, pos_out, len);
if (ret == 0) {
@@ -1600,7 +1597,7 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
if (file_out->f_op->copy_file_range) {
ret = file_out->f_op->copy_file_range(file_in, pos_in, file_out,
pos_out, len, flags);
- if (ret != -EOPNOTSUPP)
+ if (ret != -EOPNOTSUPP && ret != -EXDEV)
goto done;
}
--
1.8.3.1
From: kbuild test robot <[email protected]>
fs/nfs/nfs4file.c:345:12-18: WARNING: PTR_ERR_OR_ZERO can be used
Use PTR_ERR_OR_ZERO rather than if(IS_ERR(...)) + PTR_ERR
Generated by: scripts/coccinelle/api/ptr_ret.cocci
Fixes: 53b276379386 ("NFS inter ssc open")
CC: Olga Kornievskaia <[email protected]>
Signed-off-by: kbuild test robot <[email protected]>
---
url: https://github.com/0day-ci/linux/commits/Olga-Kornievskaia/client-side-support-for-inter-SSC-copy/20181025-043859
base: git://git.linux-nfs.org/projects/trondmy/linux-nfs.git linux-next
nfs4file.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
--- a/fs/nfs/nfs4file.c
+++ b/fs/nfs/nfs4file.c
@@ -342,7 +342,7 @@ nfs42_ssc_open(struct vfsmount *ss_mnt,
res = filep;
out:
dprintk("<-- %s error %ld filep %p r_ino %p\n",
- __func__, IS_ERR(res) ? PTR_ERR(res) : 0, res, r_ino);
+ __func__, PTR_ERR_OR_ZERO(res), res, r_ino);
return res;
out_stateowner:
Hi Olga,
Thank you for the patch! Perhaps something to improve:
[auto build test WARNING on nfs/linux-next]
[also build test WARNING on v4.19]
[cannot apply to next-20181019]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]
url: https://github.com/0day-ci/linux/commits/Olga-Kornievskaia/client-side-support-for-inter-SSC-copy/20181025-043859
base: git://git.linux-nfs.org/projects/trondmy/linux-nfs.git linux-next
coccinelle warnings: (new ones prefixed by >>)
>> fs/nfs/nfs4file.c:345:12-18: WARNING: PTR_ERR_OR_ZERO can be used
Please review and possibly fold the followup patch.
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
On Wed, Oct 24, 2018 at 10:59 PM Olga Kornievskaia
<[email protected]> wrote:
>
> From: Olga Kornievskaia <[email protected]>
>
> A proposed VFS change removes the check for the files to reside
> under the same file system. Instead, a file system driver implementation
> is allowed to perform a cross-device copy_file_range() and if
> the file system fails to support it instead fallback to doing
> a do_splice copy. Therefore, EXDEV is no longer a possible error.
>
> Signed-off-by: Olga Kornievskaia <[email protected]>
> ---
> man2/copy_file_range.2 | 7 ++-----
> 1 file changed, 2 insertions(+), 5 deletions(-)
>
> diff --git a/man2/copy_file_range.2 b/man2/copy_file_range.2
> index 20374ab..723b2d0 100644
> --- a/man2/copy_file_range.2
> +++ b/man2/copy_file_range.2
> @@ -39,7 +39,8 @@ The
> .BR copy_file_range ()
> system call performs an in-kernel copy between two file descriptors
> without the additional cost of transferring data from the kernel to user space
> -and then back into the kernel.
> +and then back into the kernel. Since kernel version 4.21(???) passed in
> +file descriptors are not required to be under the same mounted file system.
> It copies up to
> .I len
> bytes of data from file descriptor
> @@ -128,10 +129,6 @@ Out of memory.
> .B ENOSPC
> There is not enough space on the target filesystem to complete the copy.
> .TP
> -.B EXDEV
> -The files referred to by
> -.IR file_in " and " file_out
> -are not on the same mounted filesystem.
Man page serves users of old kernels as well. You should not delete this
expected error, but you can add "... and kernel does not support cross
filesystem copy".
Thanks,
Amir.
On Wed, Oct 24, 2018 at 10:58 PM Olga Kornievskaia
<[email protected]> wrote:
>
> From: Olga Kornievskaia <[email protected]>
>
> This patch removes the check for source and destination files to
> come from the same superblock. This feature was of interest to
> NFS as well as CIFS communities.
>
> Specifically, this feature is needed to allow for NFSv4.2 copy offload
> to be done between different NFSv4.2 servers. SMBv3 copy offload between
> different servers would be able to use this as well.
>
> Removal of the check implies that passed in source and destination
> files can come from different superblocks of the same file system
> type or different. It is upto each individual copy_file_range()
> file system implementation to decide what type of copy it is
> capable of doing and return -EXDEV in cases support is lacking.
>
> There are 3 known implementator of copy_file_range() f_op: NFS,
> CIFS, OverlayFS. NFS and CIFS are interested to support cross-device
> copy offload but do not support cross file system types copy offload.
> Following patches will add appropriate checks in each of the drivers.
>
That should be the other way around - first add limitation inside filesystems
then relax vfs check. otherwise you leave a bug in the middle of bisect.
> If the copy_file_range() errors with EXDEV, the code would fallback
> on doing do_splice_direct() copying which in itself is beneficial.
>
> Adding wording to the vfs.txt and porting documentation about the
> new support for cross-device copy offload.
>
> Signed-off-by: Olga Kornievskaia <[email protected]>
> ---
> Documentation/filesystems/porting | 7 +++++++
> Documentation/filesystems/vfs.txt | 6 +++++-
> fs/read_write.c | 9 +++------
> 3 files changed, 15 insertions(+), 7 deletions(-)
>
> diff --git a/Documentation/filesystems/porting b/Documentation/filesystems/porting
> index 7b7b845..ebb4954 100644
> --- a/Documentation/filesystems/porting
> +++ b/Documentation/filesystems/porting
> @@ -622,3 +622,10 @@ in your dentry operations instead.
> alloc_file_clone(file, flags, ops) does not affect any caller's references.
> On success you get a new struct file sharing the mount/dentry with the
> original, on failure - ERR_PTR().
> +--
> +[mandatory]
> + ->copy_file_range() may now be passed files which belong to two
> + different superblocks of the same file system type or which belong
> + to two different filesystems types all together. As before, the
> + destination's copy_file_range() is the function which is called.
> + If it cannot copy ranges from the source, it should return -EXDEV.
> diff --git a/Documentation/filesystems/vfs.txt b/Documentation/filesystems/vfs.txt
> index a6c6a8a..34c0e8c 100644
> --- a/Documentation/filesystems/vfs.txt
> +++ b/Documentation/filesystems/vfs.txt
> @@ -1,5 +1,6 @@
>
> Overview of the Linux Virtual File System
> +- [fs] nfs: Don't let readdirplus revalidate an inode that was marked as stale (Benjamin Coddington) [1429514 1416532]
>
???
Thanks,
Amir.
On Wed, Oct 24, 2018 at 10:59 PM Olga Kornievskaia
<[email protected]> wrote:
>
> From: Olga Kornievskaia <[email protected]>
>
> VFS copy_file_range was relaxed to allow for cross-device copy.
> Add a check to disallow cross file systems copy offload, both
> files are expected to be of NFS type.
>
> Signed-off-by: Olga Kornievskaia <[email protected]>
> ---
> fs/nfs/nfs4file.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c
> index 4288a6e..7137e7b 100644
> --- a/fs/nfs/nfs4file.c
> +++ b/fs/nfs/nfs4file.c
> @@ -135,6 +135,9 @@ static ssize_t nfs4_copy_file_range(struct file *file_in, loff_t pos_in,
> {
> ssize_t ret;
>
> + if (file_in->f_inode->i_sb->s_type != file_out->f_inode->i_sb->s_type)
> + return -EXDEV;
> +
You would need to doo the patches in following order:
1. add same sb check in nfs
2. relax same sb check in vfs
3. implement cross fs copy in nfs
4. relax same sb to same sb type in nfs
Thanks,
Amir.
On Wed, Oct 24, 2018 at 10:59 PM Olga Kornievskaia
<[email protected]> wrote:
>
> From: Olga Kornievskaia <[email protected]>
>
> VFS copy_file_range was relaxed to allow for cross-device copy.
> Add a check for now, to disallow cross-device copy offload until
> appropriate support is added.
>
> Signed-off-by: Olga Kornievskaia <[email protected]>
> `
> ---
> fs/overlayfs/file.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/fs/overlayfs/file.c b/fs/overlayfs/file.c
> index aeaefd2..5282853 100644
> --- a/fs/overlayfs/file.c
> +++ b/fs/overlayfs/file.c
> @@ -483,6 +483,9 @@ static ssize_t ovl_copy_file_range(struct file *file_in, loff_t pos_in,
> struct file *file_out, loff_t pos_out,
> size_t len, unsigned int flags)
> {
> + if (file_in->f_inode->i_sb != file_out->f_inode->i_sb)
> + return -EXDEV;
>
Note that this could be relaxed later on to same fs type
allowing for cross container copy_file_range() to be performced
as clone_file_range() when upper is a shared fs with clone support.
Don't worry about it though, I can make this change later after testing
the use case.
As long as this patch comes in the series *before* relaxing the VFS check,
You may add:
Reviewed-by: Amir Goldstein <[email protected]>
Thanks,
Amir.
On Thu, Oct 25, 2018 at 12:28 AM Amir Goldstein <[email protected]> wrote:
>
> On Wed, Oct 24, 2018 at 10:59 PM Olga Kornievskaia
> <[email protected]> wrote:
> >
> > From: Olga Kornievskaia <[email protected]>
> >
> > A proposed VFS change removes the check for the files to reside
> > under the same file system. Instead, a file system driver implementation
> > is allowed to perform a cross-device copy_file_range() and if
> > the file system fails to support it instead fallback to doing
> > a do_splice copy. Therefore, EXDEV is no longer a possible error.
> >
> > Signed-off-by: Olga Kornievskaia <[email protected]>
> > ---
> > man2/copy_file_range.2 | 7 ++-----
> > 1 file changed, 2 insertions(+), 5 deletions(-)
> >
> > diff --git a/man2/copy_file_range.2 b/man2/copy_file_range.2
> > index 20374ab..723b2d0 100644
> > --- a/man2/copy_file_range.2
> > +++ b/man2/copy_file_range.2
> > @@ -39,7 +39,8 @@ The
> > .BR copy_file_range ()
> > system call performs an in-kernel copy between two file descriptors
> > without the additional cost of transferring data from the kernel to user space
> > -and then back into the kernel.
> > +and then back into the kernel. Since kernel version 4.21(???) passed in
> > +file descriptors are not required to be under the same mounted file system.
> > It copies up to
> > .I len
> > bytes of data from file descriptor
> > @@ -128,10 +129,6 @@ Out of memory.
> > .B ENOSPC
> > There is not enough space on the target filesystem to complete the copy.
> > .TP
> > -.B EXDEV
> > -The files referred to by
> > -.IR file_in " and " file_out
> > -are not on the same mounted filesystem.
>
> Man page serves users of old kernels as well. You should not delete this
> expected error, but you can add "... and kernel does not support cross
> filesystem copy".
Thank you, will do.
>
> Thanks,
> Amir.
On Thu, Oct 25, 2018 at 12:38 AM Amir Goldstein <[email protected]> wrote:
>
> On Wed, Oct 24, 2018 at 10:59 PM Olga Kornievskaia
> <[email protected]> wrote:
> >
> > From: Olga Kornievskaia <[email protected]>
> >
> > VFS copy_file_range was relaxed to allow for cross-device copy.
> > Add a check to disallow cross file systems copy offload, both
> > files are expected to be of NFS type.
> >
> > Signed-off-by: Olga Kornievskaia <[email protected]>
> > ---
> > fs/nfs/nfs4file.c | 3 +++
> > 1 file changed, 3 insertions(+)
> >
> > diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c
> > index 4288a6e..7137e7b 100644
> > --- a/fs/nfs/nfs4file.c
> > +++ b/fs/nfs/nfs4file.c
> > @@ -135,6 +135,9 @@ static ssize_t nfs4_copy_file_range(struct file *file_in, loff_t pos_in,
> > {
> > ssize_t ret;
> >
> > + if (file_in->f_inode->i_sb->s_type != file_out->f_inode->i_sb->s_type)
> > + return -EXDEV;
> > +
>
> You would need to doo the patches in following order:
> 1. add same sb check in nfs
> 2. relax same sb check in vfs
> 3. implement cross fs copy in nfs
> 4. relax same sb to same sb type in nfs
Thank you, I will reorder the patches.
>
> Thanks,
> Amir.
On Thu, Oct 25, 2018 at 1:54 AM Amir Goldstein <[email protected]> wrote:
>
> On Wed, Oct 24, 2018 at 10:59 PM Olga Kornievskaia
> <[email protected]> wrote:
> >
> > From: Olga Kornievskaia <[email protected]>
> >
> > VFS copy_file_range was relaxed to allow for cross-device copy.
> > Add a check for now, to disallow cross-device copy offload until
> > appropriate support is added.
> >
> > Signed-off-by: Olga Kornievskaia <[email protected]>
> > `
> > ---
> > fs/overlayfs/file.c | 3 +++
> > 1 file changed, 3 insertions(+)
> >
> > diff --git a/fs/overlayfs/file.c b/fs/overlayfs/file.c
> > index aeaefd2..5282853 100644
> > --- a/fs/overlayfs/file.c
> > +++ b/fs/overlayfs/file.c
> > @@ -483,6 +483,9 @@ static ssize_t ovl_copy_file_range(struct file *file_in, loff_t pos_in,
> > struct file *file_out, loff_t pos_out,
> > size_t len, unsigned int flags)
> > {
> > + if (file_in->f_inode->i_sb != file_out->f_inode->i_sb)
> > + return -EXDEV;
> >
>
> Note that this could be relaxed later on to same fs type
> allowing for cross container copy_file_range() to be performced
> as clone_file_range() when upper is a shared fs with clone support.
>
> Don't worry about it though, I can make this change later after testing
> the use case.
>
> As long as this patch comes in the series *before* relaxing the VFS check,
> You may add:
> Reviewed-by: Amir Goldstein <[email protected]>
Thank you. I will reorder the patches and yes I thought the individual
FS maintainers/folks would relax the check according to their
functionality. This patch just provides same semantics with respect to
the cross device checking but at the driver level.
>
> Thanks,
> Amir.
On Thu, Oct 25, 2018 at 6:28 PM Olga Kornievskaia
<[email protected]> wrote:
>
> On Thu, Oct 25, 2018 at 12:38 AM Amir Goldstein <[email protected]> wrote:
> >
> > On Wed, Oct 24, 2018 at 10:59 PM Olga Kornievskaia
> > <[email protected]> wrote:
> > >
> > > From: Olga Kornievskaia <[email protected]>
> > >
> > > VFS copy_file_range was relaxed to allow for cross-device copy.
> > > Add a check to disallow cross file systems copy offload, both
> > > files are expected to be of NFS type.
> > >
> > > Signed-off-by: Olga Kornievskaia <[email protected]>
> > > ---
> > > fs/nfs/nfs4file.c | 3 +++
> > > 1 file changed, 3 insertions(+)
> > >
> > > diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c
> > > index 4288a6e..7137e7b 100644
> > > --- a/fs/nfs/nfs4file.c
> > > +++ b/fs/nfs/nfs4file.c
> > > @@ -135,6 +135,9 @@ static ssize_t nfs4_copy_file_range(struct file *file_in, loff_t pos_in,
> > > {
> > > ssize_t ret;
> > >
> > > + if (file_in->f_inode->i_sb->s_type != file_out->f_inode->i_sb->s_type)
> > > + return -EXDEV;
> > > +
> >
> > You would need to doo the patches in following order:
> > 1. add same sb check in nfs
> > 2. relax same sb check in vfs
> > 3. implement cross fs copy in nfs
> > 4. relax same sb to same sb type in nfs
>
> Thank you, I will reorder the patches.
>
On second thought, It's probably best to do #1 (for nfs, cifs, ovl) + #2
in the same patch, to prevent someone from backporting just #2
without #1.
Let me know if this was too cryptic.
Thanks,
Amir.
On Thu, Oct 25, 2018 at 12:54 PM Amir Goldstein <[email protected]> wrote:
>
> On Thu, Oct 25, 2018 at 6:28 PM Olga Kornievskaia
> <[email protected]> wrote:
> >
> > On Thu, Oct 25, 2018 at 12:38 AM Amir Goldstein <[email protected]> wrote:
> > >
> > > On Wed, Oct 24, 2018 at 10:59 PM Olga Kornievskaia
> > > <[email protected]> wrote:
> > > >
> > > > From: Olga Kornievskaia <[email protected]>
> > > >
> > > > VFS copy_file_range was relaxed to allow for cross-device copy.
> > > > Add a check to disallow cross file systems copy offload, both
> > > > files are expected to be of NFS type.
> > > >
> > > > Signed-off-by: Olga Kornievskaia <[email protected]>
> > > > ---
> > > > fs/nfs/nfs4file.c | 3 +++
> > > > 1 file changed, 3 insertions(+)
> > > >
> > > > diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c
> > > > index 4288a6e..7137e7b 100644
> > > > --- a/fs/nfs/nfs4file.c
> > > > +++ b/fs/nfs/nfs4file.c
> > > > @@ -135,6 +135,9 @@ static ssize_t nfs4_copy_file_range(struct file *file_in, loff_t pos_in,
> > > > {
> > > > ssize_t ret;
> > > >
> > > > + if (file_in->f_inode->i_sb->s_type != file_out->f_inode->i_sb->s_type)
> > > > + return -EXDEV;
> > > > +
> > >
> > > You would need to doo the patches in following order:
> > > 1. add same sb check in nfs
> > > 2. relax same sb check in vfs
> > > 3. implement cross fs copy in nfs
> > > 4. relax same sb to same sb type in nfs
> >
> > Thank you, I will reorder the patches.
> >
>
> On second thought, It's probably best to do #1 (for nfs, cifs, ovl) + #2
> in the same patch, to prevent someone from backporting just #2
> without #1.
>
> Let me know if this was too cryptic.
Thank you I think I understood.
In an unsubmitted version I had a single VFS patch that made the VFS
change and included the checks for the file systems. However, I
thought maybe it wasn't appropriate to have FS changes and VFS changes
in the same patch and thus I split them up. Since you say it's
actually preferred I will do so. Do I need to change the title from
just "VFS" to somehow indicate that it has NFS/CIFS/Overlayfs content?
Or is the current title ok?
>
> Thanks,
> Amir.
On Thu, Oct 25, 2018 at 07:28:10AM +0300, Amir Goldstein wrote:
> On Wed, Oct 24, 2018 at 10:59 PM Olga Kornievskaia
> <[email protected]> wrote:
> >
> > From: Olga Kornievskaia <[email protected]>
> >
> > A proposed VFS change removes the check for the files to reside
> > under the same file system. Instead, a file system driver implementation
> > is allowed to perform a cross-device copy_file_range() and if
> > the file system fails to support it instead fallback to doing
> > a do_splice copy. Therefore, EXDEV is no longer a possible error.
> >
> > Signed-off-by: Olga Kornievskaia <[email protected]>
> > ---
> > man2/copy_file_range.2 | 7 ++-----
> > 1 file changed, 2 insertions(+), 5 deletions(-)
> >
> > diff --git a/man2/copy_file_range.2 b/man2/copy_file_range.2
> > index 20374ab..723b2d0 100644
> > --- a/man2/copy_file_range.2
> > +++ b/man2/copy_file_range.2
> > @@ -39,7 +39,8 @@ The
> > .BR copy_file_range ()
> > system call performs an in-kernel copy between two file descriptors
> > without the additional cost of transferring data from the kernel to user space
> > -and then back into the kernel.
> > +and then back into the kernel. Since kernel version 4.21(???) passed in
> > +file descriptors are not required to be under the same mounted file system.
> > It copies up to
> > .I len
> > bytes of data from file descriptor
> > @@ -128,10 +129,6 @@ Out of memory.
> > .B ENOSPC
> > There is not enough space on the target filesystem to complete the copy.
> > .TP
> > -.B EXDEV
> > -The files referred to by
> > -.IR file_in " and " file_out
> > -are not on the same mounted filesystem.
>
> Man page serves users of old kernels as well. You should not delete this
> expected error, but you can add "... and kernel does not support cross
> filesystem copy".
I'd rather see this worded as:
.B EXDEV
The files referred to by
.IR file_in " and " file_out
-are not on the same mounted filesystem.
+cannot be copied due to being on incompatible filesystems
On Thu, Oct 25, 2018 at 1:24 PM Matthew Wilcox <[email protected]> wrote:
>
> On Thu, Oct 25, 2018 at 07:28:10AM +0300, Amir Goldstein wrote:
> > On Wed, Oct 24, 2018 at 10:59 PM Olga Kornievskaia
> > <[email protected]> wrote:
> > >
> > > From: Olga Kornievskaia <[email protected]>
> > >
> > > A proposed VFS change removes the check for the files to reside
> > > under the same file system. Instead, a file system driver implementation
> > > is allowed to perform a cross-device copy_file_range() and if
> > > the file system fails to support it instead fallback to doing
> > > a do_splice copy. Therefore, EXDEV is no longer a possible error.
> > >
> > > Signed-off-by: Olga Kornievskaia <[email protected]>
> > > ---
> > > man2/copy_file_range.2 | 7 ++-----
> > > 1 file changed, 2 insertions(+), 5 deletions(-)
> > >
> > > diff --git a/man2/copy_file_range.2 b/man2/copy_file_range.2
> > > index 20374ab..723b2d0 100644
> > > --- a/man2/copy_file_range.2
> > > +++ b/man2/copy_file_range.2
> > > @@ -39,7 +39,8 @@ The
> > > .BR copy_file_range ()
> > > system call performs an in-kernel copy between two file descriptors
> > > without the additional cost of transferring data from the kernel to user space
> > > -and then back into the kernel.
> > > +and then back into the kernel. Since kernel version 4.21(???) passed in
> > > +file descriptors are not required to be under the same mounted file system.
> > > It copies up to
> > > .I len
> > > bytes of data from file descriptor
> > > @@ -128,10 +129,6 @@ Out of memory.
> > > .B ENOSPC
> > > There is not enough space on the target filesystem to complete the copy.
> > > .TP
> > > -.B EXDEV
> > > -The files referred to by
> > > -.IR file_in " and " file_out
> > > -are not on the same mounted filesystem.
> >
> > Man page serves users of old kernels as well. You should not delete this
> > expected error, but you can add "... and kernel does not support cross
> > filesystem copy".
>
> I'd rather see this worded as:
>
> .B EXDEV
> The files referred to by
> .IR file_in " and " file_out
> -are not on the same mounted filesystem.
> +cannot be copied due to being on incompatible filesystems
Why should this be changed to "incompatible filesystems". Existing
code doesn't support not only incompatibly filesystem but also cross
device of compatible file systems. And this error is only left for the
old kernels. In the new kernel, the error EXDEV will never be
returned.
On Thu, Oct 25, 2018 at 01:47:21PM -0400, Olga Kornievskaia wrote:
> On Thu, Oct 25, 2018 at 1:24 PM Matthew Wilcox <[email protected]> wrote:
> >
> > On Thu, Oct 25, 2018 at 07:28:10AM +0300, Amir Goldstein wrote:
> > > On Wed, Oct 24, 2018 at 10:59 PM Olga Kornievskaia
> > > <[email protected]> wrote:
> > > >
> > > > From: Olga Kornievskaia <[email protected]>
> > > >
> > > > A proposed VFS change removes the check for the files to reside
> > > > under the same file system. Instead, a file system driver implementation
> > > > is allowed to perform a cross-device copy_file_range() and if
> > > > the file system fails to support it instead fallback to doing
> > > > a do_splice copy. Therefore, EXDEV is no longer a possible error.
> > > >
> > > > Signed-off-by: Olga Kornievskaia <[email protected]>
> > > > ---
> > > > man2/copy_file_range.2 | 7 ++-----
> > > > 1 file changed, 2 insertions(+), 5 deletions(-)
> > > >
> > > > diff --git a/man2/copy_file_range.2 b/man2/copy_file_range.2
> > > > index 20374ab..723b2d0 100644
> > > > --- a/man2/copy_file_range.2
> > > > +++ b/man2/copy_file_range.2
> > > > @@ -39,7 +39,8 @@ The
> > > > .BR copy_file_range ()
> > > > system call performs an in-kernel copy between two file descriptors
> > > > without the additional cost of transferring data from the kernel to user space
> > > > -and then back into the kernel.
> > > > +and then back into the kernel. Since kernel version 4.21(???) passed in
> > > > +file descriptors are not required to be under the same mounted file system.
> > > > It copies up to
> > > > .I len
> > > > bytes of data from file descriptor
> > > > @@ -128,10 +129,6 @@ Out of memory.
> > > > .B ENOSPC
> > > > There is not enough space on the target filesystem to complete the copy.
> > > > .TP
> > > > -.B EXDEV
> > > > -The files referred to by
> > > > -.IR file_in " and " file_out
> > > > -are not on the same mounted filesystem.
> > >
> > > Man page serves users of old kernels as well. You should not delete this
> > > expected error, but you can add "... and kernel does not support cross
> > > filesystem copy".
> >
> > I'd rather see this worded as:
> >
> > .B EXDEV
> > The files referred to by
> > .IR file_in " and " file_out
> > -are not on the same mounted filesystem.
> > +cannot be copied due to being on incompatible filesystems
>
> Why should this be changed to "incompatible filesystems". Existing
> code doesn't support not only incompatibly filesystem but also cross
> device of compatible file systems. And this error is only left for the
> old kernels. In the new kernel, the error EXDEV will never be
> returned.
If I try to copy_file_range() between an NFS filesystem and a CIFS
filesystem with a new kernel, what error do I get?
On Thu, Oct 25, 2018 at 2:08 PM Matthew Wilcox <[email protected]> wrote:
>
> On Thu, Oct 25, 2018 at 01:47:21PM -0400, Olga Kornievskaia wrote:
> > On Thu, Oct 25, 2018 at 1:24 PM Matthew Wilcox <[email protected]> wrote:
> > >
> > > On Thu, Oct 25, 2018 at 07:28:10AM +0300, Amir Goldstein wrote:
> > > > On Wed, Oct 24, 2018 at 10:59 PM Olga Kornievskaia
> > > > <[email protected]> wrote:
> > > > >
> > > > > From: Olga Kornievskaia <[email protected]>
> > > > >
> > > > > A proposed VFS change removes the check for the files to reside
> > > > > under the same file system. Instead, a file system driver implementation
> > > > > is allowed to perform a cross-device copy_file_range() and if
> > > > > the file system fails to support it instead fallback to doing
> > > > > a do_splice copy. Therefore, EXDEV is no longer a possible error.
> > > > >
> > > > > Signed-off-by: Olga Kornievskaia <[email protected]>
> > > > > ---
> > > > > man2/copy_file_range.2 | 7 ++-----
> > > > > 1 file changed, 2 insertions(+), 5 deletions(-)
> > > > >
> > > > > diff --git a/man2/copy_file_range.2 b/man2/copy_file_range.2
> > > > > index 20374ab..723b2d0 100644
> > > > > --- a/man2/copy_file_range.2
> > > > > +++ b/man2/copy_file_range.2
> > > > > @@ -39,7 +39,8 @@ The
> > > > > .BR copy_file_range ()
> > > > > system call performs an in-kernel copy between two file descriptors
> > > > > without the additional cost of transferring data from the kernel to user space
> > > > > -and then back into the kernel.
> > > > > +and then back into the kernel. Since kernel version 4.21(???) passed in
> > > > > +file descriptors are not required to be under the same mounted file system.
> > > > > It copies up to
> > > > > .I len
> > > > > bytes of data from file descriptor
> > > > > @@ -128,10 +129,6 @@ Out of memory.
> > > > > .B ENOSPC
> > > > > There is not enough space on the target filesystem to complete the copy.
> > > > > .TP
> > > > > -.B EXDEV
> > > > > -The files referred to by
> > > > > -.IR file_in " and " file_out
> > > > > -are not on the same mounted filesystem.
> > > >
> > > > Man page serves users of old kernels as well. You should not delete this
> > > > expected error, but you can add "... and kernel does not support cross
> > > > filesystem copy".
> > >
> > > I'd rather see this worded as:
> > >
> > > .B EXDEV
> > > The files referred to by
> > > .IR file_in " and " file_out
> > > -are not on the same mounted filesystem.
> > > +cannot be copied due to being on incompatible filesystems
> >
> > Why should this be changed to "incompatible filesystems". Existing
> > code doesn't support not only incompatibly filesystem but also cross
> > device of compatible file systems. And this error is only left for the
> > old kernels. In the new kernel, the error EXDEV will never be
> > returned.
>
> If I try to copy_file_range() between an NFS filesystem and a CIFS
> filesystem with a new kernel, what error do I get?
It is my understanding that you will not get an error instead a
do_splice_direct() will do that copy.
On Thu, Oct 25, 2018 at 1:54 AM Amir Goldstein <[email protected]> wrote:
>
> On Wed, Oct 24, 2018 at 10:59 PM Olga Kornievskaia
> <[email protected]> wrote:
> >
> > From: Olga Kornievskaia <[email protected]>
> >
> > VFS copy_file_range was relaxed to allow for cross-device copy.
> > Add a check for now, to disallow cross-device copy offload until
> > appropriate support is added.
> >
> > Signed-off-by: Olga Kornievskaia <[email protected]>
> > `
> > ---
> > fs/overlayfs/file.c | 3 +++
> > 1 file changed, 3 insertions(+)
> >
> > diff --git a/fs/overlayfs/file.c b/fs/overlayfs/file.c
> > index aeaefd2..5282853 100644
> > --- a/fs/overlayfs/file.c
> > +++ b/fs/overlayfs/file.c
> > @@ -483,6 +483,9 @@ static ssize_t ovl_copy_file_range(struct file *file_in, loff_t pos_in,
> > struct file *file_out, loff_t pos_out,
> > size_t len, unsigned int flags)
> > {
> > + if (file_in->f_inode->i_sb != file_out->f_inode->i_sb)
> > + return -EXDEV;
> >
>
> Note that this could be relaxed later on to same fs type
> allowing for cross container copy_file_range() to be performced
> as clone_file_range() when upper is a shared fs with clone support.
>
> Don't worry about it though, I can make this change later after testing
> the use case.
>
> As long as this patch comes in the series *before* relaxing the VFS check,
> You may add:
> Reviewed-by: Amir Goldstein <[email protected]>
Given that this patch was merged I didn't add "reviewed-by" line.
Please let me know I need to add this line to the main VFS patch.
>
> Thanks,
> Amir.