Skip to contentSkip to conversation

nfs/nfsd: add support for localio

PATCH Superseded Mike Snitzer 4 participants 55 messages Aug 23, 2024 – Aug 29, 2024

Subsystems (5)

v13 of 16 Superseded Latest is v16 →

lore
Share & tools
lore
#0 lore raw

v13 cover nfs/nfsd: add support for localio

These latest changes are available in my git tree here:
https://git.kernel.org/pub/scm/linux/kernel/git/snitzer/linux.git/log/?h=nfs-localio-for-next

Changes since v12:
- Rebased to rearrange code to avoid "churn" that Jeff Layton felt distracting (biggest improvement came from folding switch from struct file to nfsd_file changes in from the start of the series)
- Updated relevant patch headers accordingly.
- Updated localio.rst to provide more performance data.
- Dropped v12's buggy "nfsd: fix nfsfh tracepoints to properly handle NULL rqstp" patch -- fixing localio to support fh_verify tracepoints will need more think-time and discussion, but they aren't of critical importance so fixing them doesn't need to hold things up.

Please also see v12's patch header for v12's extensive changes:
https://marc.info/?l=linux-nfs&m=172409139907981&w=2

Thanks,
Mike

Mike Snitzer (9):
nfs_common: factor out nfs_errtbl and nfs_stat_to_errno
nfs_common: factor out nfs4_errtbl and nfs4_stat_to_errno
nfs: factor out {encode,decode}_opaque_fixed to nfs_xdr.h
SUNRPC: remove call_allocate() BUG_ONs
nfs_common: add NFS LOCALIO auxiliary protocol enablement
nfsd: implement server support for NFS_LOCALIO_PROGRAM
nfs: pass struct nfsd_file to nfs_init_pgio and nfs_init_commit
nfs: implement client support for NFS_LOCALIO_PROGRAM
nfs: add Documentation/filesystems/nfs/localio.rst

NeilBrown (3):
nfsd: factor out __fh_verify to allow NULL rqstp to be passed
nfsd: add nfsd_file_acquire_local()
SUNRPC: replace program list with program array

Trond Myklebust (4):
nfs: enable localio for non-pNFS IO
pnfs/flexfiles: enable localio support
nfs/localio: use dedicated workqueues for filesystem read and write
nfs: add FAQ section to Documentation/filesystems/nfs/localio.rst

Weston Andros Adamson (3):
SUNRPC: add rpcauth_map_clnt_to_svc_cred_local
nfsd: add localio support
nfs: add localio support

 Documentation/filesystems/nfs/localio.rst | 276 ++++++++
 fs/Kconfig                                |   3 +
 fs/nfs/Kconfig                            |  15 +
 fs/nfs/Makefile                           |   1 +
 fs/nfs/client.c                           |  15 +-
 fs/nfs/filelayout/filelayout.c            |   6 +-
 fs/nfs/flexfilelayout/flexfilelayout.c    | 142 +++-
 fs/nfs/flexfilelayout/flexfilelayout.h    |   2 +
 fs/nfs/flexfilelayout/flexfilelayoutdev.c |   6 +
 fs/nfs/inode.c                            |  57 +-
 fs/nfs/internal.h                         |  61 +-
 fs/nfs/localio.c                          | 784 ++++++++++++++++++++++
 fs/nfs/nfs2xdr.c                          |  70 +-
 fs/nfs/nfs3xdr.c                          | 108 +--
 fs/nfs/nfs4xdr.c                          |  84 +--
 fs/nfs/nfstrace.h                         |  61 ++
 fs/nfs/pagelist.c                         |  16 +-
 fs/nfs/pnfs_nfs.c                         |   2 +-
 fs/nfs/write.c                            |  12 +-
 fs/nfs_common/Makefile                    |   5 +
 fs/nfs_common/common.c                    | 134 ++++
 fs/nfs_common/nfslocalio.c                | 194 ++++++
 fs/nfsd/Kconfig                           |  15 +
 fs/nfsd/Makefile                          |   1 +
 fs/nfsd/export.c                          |   8 +-
 fs/nfsd/filecache.c                       |  90 ++-
 fs/nfsd/filecache.h                       |   5 +
 fs/nfsd/localio.c                         | 183 +++++
 fs/nfsd/netns.h                           |   8 +-
 fs/nfsd/nfsctl.c                          |   2 +-
 fs/nfsd/nfsd.h                            |   6 +-
 fs/nfsd/nfsfh.c                           | 122 ++--
 fs/nfsd/nfsfh.h                           |   5 +
 fs/nfsd/nfssvc.c                          | 100 ++-
 fs/nfsd/trace.h                           |   3 +-
 fs/nfsd/vfs.h                             |  10 +
 include/linux/nfs.h                       |   9 +
 include/linux/nfs_common.h                |  17 +
 include/linux/nfs_fs_sb.h                 |  10 +
 include/linux/nfs_xdr.h                   |  20 +-
 include/linux/nfslocalio.h                |  56 ++
 include/linux/sunrpc/auth.h               |   4 +
 include/linux/sunrpc/svc.h                |   7 +-
 net/sunrpc/auth.c                         |  22 +
 net/sunrpc/clnt.c                         |   6 -
 net/sunrpc/svc.c                          |  68 +-
 net/sunrpc/svc_xprt.c                     |   2 +-
 net/sunrpc/svcauth_unix.c                 |   3 +-
 48 files changed, 2434 insertions(+), 402 deletions(-)
 create mode 100644 Documentation/filesystems/nfs/localio.rst
 create mode 100644 fs/nfs/localio.c
 create mode 100644 fs/nfs_common/common.c
 create mode 100644 fs/nfs_common/nfslocalio.c
 create mode 100644 fs/nfsd/localio.c
 create mode 100644 include/linux/nfs_common.h
 create mode 100644 include/linux/nfslocalio.h
Signature · 1 line
2.44.0
#1 lore raw

PATCH v13 1/19 nfs_common: factor out nfs_errtbl and nfs_stat_to_errno

Common nfs_stat_to_errno() is used by both fs/nfs/nfs2xdr.c and fs/nfs/nfs3xdr.c

Will also be used by fs/nfsd/localio.c

---
 fs/nfs/Kconfig             |   1 +
 fs/nfs/nfs2xdr.c           |  70 +-----------------------
 fs/nfs/nfs3xdr.c           | 108 +++++++------------------------------
 fs/nfs/nfs4xdr.c           |   4 +-
 fs/nfs_common/Makefile     |   2 +
 fs/nfs_common/common.c     |  67 +++++++++++++++++++++++
 fs/nfsd/Kconfig            |   1 +
 include/linux/nfs_common.h |  16 ++++++
 8 files changed, 109 insertions(+), 160 deletions(-)
 create mode 100644 fs/nfs_common/common.c
 create mode 100644 include/linux/nfs_common.h
fs/nfs/Kconfig +1 -0
index 57249f040dfc..0eb20012792f 100644
@@ -4,6 +4,7 @@ config NFS_FS
 	depends on INET && FILE_LOCKING && MULTIUSER
 	select LOCKD
 	select SUNRPC
+	select NFS_COMMON
 	select NFS_ACL_SUPPORT if NFS_V3_ACL
 	help
 	  Choose Y here if you want to access files residing on other
fs/nfs/nfs2xdr.c +1 -69

93 lines not on this page · raw message

fs/nfs/nfs3xdr.c +20 -88

237 lines not on this page · raw message

fs/nfs/nfs4xdr.c +1 -3
index 7704a4509676..b4091af1a60d 100644
@@ -52,6 +52,7 @@
 #include <linux/nfs.h>
 #include <linux/nfs4.h>
 #include <linux/nfs_fs.h>
+#include <linux/nfs_common.h>
 
 #include "nfs4_fs.h"
 #include "nfs4trace.h"
@@ -63,9 +64,6 @@
 
 #define NFSDBG_FACILITY		NFSDBG_XDR
 
-/* Mapping from NFS error code to "errno" error code. */
-#define errno_NFSERR_IO		EIO
-
 struct compound_hdr;
 static int nfs4_stat_to_errno(int);
 static void encode_layoutget(struct xdr_stream *xdr,
fs/nfs_common/Makefile +2 -0
index 119c75ab9fd0..e58b01bb8dda 100644
@@ -8,3 +8,5 @@ nfs_acl-objs := nfsacl.o
 
 obj-$(CONFIG_GRACE_PERIOD) += grace.o
 obj-$(CONFIG_NFS_V4_2_SSC_HELPER) += nfs_ssc.o
+
+obj-$(CONFIG_NFS_COMMON) += common.o
fs/nfs_common/common.c new file +67 -0

67 lines not on this page · raw message

fs/nfsd/Kconfig +1 -0
index ec2ab6429e00..c0bd1509ccd4 100644
@@ -7,6 +7,7 @@ config NFSD
 	select LOCKD
 	select SUNRPC
 	select EXPORTFS
+	select NFS_COMMON
 	select NFS_ACL_SUPPORT if NFSD_V2_ACL
 	select NFS_ACL_SUPPORT if NFSD_V3_ACL
 	depends on MULTIUSER
include/linux/nfs_common.h new file +16 -0
new file mode 100644
index 000000000000..3395c4a4d372
@@ -0,0 +1,16 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * This file contains constants and methods used by both NFS client and server.
+ */
+#ifndef _LINUX_NFS_COMMON_H
+#define _LINUX_NFS_COMMON_H
+
+#include <linux/errno.h>
+#include <uapi/linux/nfs.h>
+
+/* Mapping from NFS error code to "errno" error code. */
+#define errno_NFSERR_IO EIO
+
+int nfs_stat_to_errno(enum nfs_stat status);
+
+#endif /* _LINUX_NFS_COMMON_H */
Signature · 1 line
2.44.0
#2 lore raw

PATCH v13 2/19 nfs_common: factor out nfs4_errtbl and nfs4_stat_to_errno

In reply to Mike Snitzer

Common nfs4_stat_to_errno() is used by fs/nfs/nfs4xdr.c and will be used by fs/nfs/localio.c

---
 fs/nfs/nfs4xdr.c           | 67 --------------------------------------
 fs/nfs_common/common.c     | 67 ++++++++++++++++++++++++++++++++++++++
 include/linux/nfs_common.h |  1 +
 3 files changed, 68 insertions(+), 67 deletions(-)
fs/nfs/nfs4xdr.c +0 -67
index b4091af1a60d..971305bdaecb 100644
@@ -65,7 +65,6 @@
 #define NFSDBG_FACILITY		NFSDBG_XDR
 
 struct compound_hdr;
-static int nfs4_stat_to_errno(int);
 static void encode_layoutget(struct xdr_stream *xdr,
 			     const struct nfs4_layoutget_args *args,
 			     struct compound_hdr *hdr);
@@ -7619,72 +7618,6 @@ int nfs4_decode_dirent(struct xdr_stream *xdr, struct nfs_entry *entry,
 	return 0;
 }
 
-/*
- * We need to translate between nfs status return values and
- * the local errno values which may not be the same.
- */
-static struct {
-	int stat;
-	int errno;
-} nfs_errtbl[] = {
-	{ NFS4_OK,		0		},
-	{ NFS4ERR_PERM,		-EPERM		},
-	{ NFS4ERR_NOENT,	-ENOENT		},
-	{ NFS4ERR_IO,		-errno_NFSERR_IO},
-	{ NFS4ERR_NXIO,		-ENXIO		},
-	{ NFS4ERR_ACCESS,	-EACCES		},
-	{ NFS4ERR_EXIST,	-EEXIST		},
-	{ NFS4ERR_XDEV,		-EXDEV		},
-	{ NFS4ERR_NOTDIR,	-ENOTDIR	},
-	{ NFS4ERR_ISDIR,	-EISDIR		},
-	{ NFS4ERR_INVAL,	-EINVAL		},
-	{ NFS4ERR_FBIG,		-EFBIG		},
-	{ NFS4ERR_NOSPC,	-ENOSPC		},
-	{ NFS4ERR_ROFS,		-EROFS		},
-	{ NFS4ERR_MLINK,	-EMLINK		},
-	{ NFS4ERR_NAMETOOLONG,	-ENAMETOOLONG	},
-	{ NFS4ERR_NOTEMPTY,	-ENOTEMPTY	},
-	{ NFS4ERR_DQUOT,	-EDQUOT		},
-	{ NFS4ERR_STALE,	-ESTALE		},
-	{ NFS4ERR_BADHANDLE,	-EBADHANDLE	},
-	{ NFS4ERR_BAD_COOKIE,	-EBADCOOKIE	},
-	{ NFS4ERR_NOTSUPP,	-ENOTSUPP	},
-	{ NFS4ERR_TOOSMALL,	-ETOOSMALL	},
-	{ NFS4ERR_SERVERFAULT,	-EREMOTEIO	},
-	{ NFS4ERR_BADTYPE,	-EBADTYPE	},
-	{ NFS4ERR_LOCKED,	-EAGAIN		},
-	{ NFS4ERR_SYMLINK,	-ELOOP		},
-	{ NFS4ERR_OP_ILLEGAL,	-EOPNOTSUPP	},
-	{ NFS4ERR_DEADLOCK,	-EDEADLK	},
-	{ NFS4ERR_NOXATTR,	-ENODATA	},
-	{ NFS4ERR_XATTR2BIG,	-E2BIG		},
-	{ -1,			-EIO		}
-};
-
-/*
- * Convert an NFS error code to a local one.
- * This one is used jointly by NFSv2 and NFSv3.
- */
-static int
-nfs4_stat_to_errno(int stat)
-{
-	int i;
-	for (i = 0; nfs_errtbl[i].stat != -1; i++) {
-		if (nfs_errtbl[i].stat == stat)
-			return nfs_errtbl[i].errno;
-	}
-	if (stat <= 10000 || stat > 10100) {
-		/* The server is looney tunes. */
-		return -EREMOTEIO;
-	}
-	/* If we cannot translate the error, the recovery routines should
-	 * handle it.
-	 * Note: remaining NFSv4 error codes have values > 10000, so should
-	 * not conflict with native Linux error codes.
-	 */
-	return -stat;
-}
-
 #ifdef CONFIG_NFS_V4_2
 #include "nfs42xdr.c"
 #endif /* CONFIG_NFS_V4_2 */
fs/nfs_common/common.c +67 -0

76 lines not on this page · raw message

include/linux/nfs_common.h +1 -0
index 3395c4a4d372..5fc02df88252 100644
@@ -12,5 +12,6 @@
 #define errno_NFSERR_IO EIO
 
 int nfs_stat_to_errno(enum nfs_stat status);
+int nfs4_stat_to_errno(int stat);
 
 #endif /* _LINUX_NFS_COMMON_H */
Signature · 1 line
2.44.0
#3 lore raw

PATCH v13 3/19 nfs: factor out {encode,decode}_opaque_fixed to nfs_xdr.h

In reply to Mike Snitzer

Eliminates duplicate functions in various files to allow for additional callers.

---
 fs/nfs/flexfilelayout/flexfilelayout.c |  6 ------
 fs/nfs/nfs4xdr.c                       | 13 -------------
 include/linux/nfs_xdr.h                | 20 +++++++++++++++++++-
 3 files changed, 19 insertions(+), 20 deletions(-)
fs/nfs/flexfilelayout/flexfilelayout.c +0 -6
index 39ba9f4208aa..d4d551ffea7b 100644
@@ -2086,12 +2086,6 @@ static int ff_layout_encode_ioerr(struct xdr_stream *xdr,
 	return ff_layout_encode_ds_ioerr(xdr, &ff_args->errors);
 }
 
-static void
-encode_opaque_fixed(struct xdr_stream *xdr, const void *buf, size_t len)
-{
-	WARN_ON_ONCE(xdr_stream_encode_opaque_fixed(xdr, buf, len) < 0);
-}
-
 static void
 ff_layout_encode_ff_iostat_head(struct xdr_stream *xdr,
 			    const nfs4_stateid *stateid,
fs/nfs/nfs4xdr.c +0 -13
index 971305bdaecb..6bf2d44e5d4e 100644
@@ -972,11 +972,6 @@ static __be32 *reserve_space(struct xdr_stream *xdr, size_t nbytes)
 	return p;
 }
 
-static void encode_opaque_fixed(struct xdr_stream *xdr, const void *buf, size_t len)
-{
-	WARN_ON_ONCE(xdr_stream_encode_opaque_fixed(xdr, buf, len) < 0);
-}
-
 static void encode_string(struct xdr_stream *xdr, unsigned int len, const char *str)
 {
 	WARN_ON_ONCE(xdr_stream_encode_opaque(xdr, str, len) < 0);
@@ -4406,14 +4401,6 @@ static int decode_access(struct xdr_stream *xdr, u32 *supported, u32 *access)
 	return 0;
 }
 
-static int decode_opaque_fixed(struct xdr_stream *xdr, void *buf, size_t len)
-{
-	ssize_t ret = xdr_stream_decode_opaque_fixed(xdr, buf, len);
-	if (unlikely(ret < 0))
-		return -EIO;
-	return 0;
-}
-
 static int decode_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid)
 {
 	return decode_opaque_fixed(xdr, stateid, NFS4_STATEID_SIZE);
include/linux/nfs_xdr.h +19 -1
index 45623af3e7b8..5e93fbfb785a 100644
@@ -1853,6 +1853,24 @@ struct nfs_rpc_ops {
 	void	(*disable_swap)(struct inode *inode);
 };
 
+/*
+ * Helper functions used by NFS client and/or server
+ */
+static inline void encode_opaque_fixed(struct xdr_stream *xdr,
+				       const void *buf, size_t len)
+{
+	WARN_ON_ONCE(xdr_stream_encode_opaque_fixed(xdr, buf, len) < 0);
+}
+
+static inline int decode_opaque_fixed(struct xdr_stream *xdr,
+				      void *buf, size_t len)
+{
+	ssize_t ret = xdr_stream_decode_opaque_fixed(xdr, buf, len);
+	if (unlikely(ret < 0))
+		return -EIO;
+	return 0;
+}
+
 /*
  * Function vectors etc. for the NFS client
  */
@@ -1866,4 +1884,4 @@ extern const struct rpc_version nfs_version4;
 extern const struct rpc_version nfsacl_version3;
 extern const struct rpc_program nfsacl_program;
 
-#endif
+#endif /* _LINUX_NFS_XDR_H */
Signature · 1 line
2.44.0
#4 lore raw

PATCH v13 4/19 nfsd: factor out __fh_verify to allow NULL rqstp to be passed

In reply to Mike Snitzer

From: NeilBrown <[email protected]>

__fh_verify() offers an interface like fh_verify() but doesn't require a struct svc_rqst *, instead it also takes the specific parts as explicit required arguments. So it is safe to call __fh_verify() with a NULL rqstp, but the net, cred, and client args must not be NULL.

__fh_verify() does not use SVC_NET(), nor does the functions it calls.

Rather then depending on rqstp->rq_vers to determine nfs version, pass it in explicitly. This removes another dependency on rqstp and ensures the correct version is checked. The rqstp can be for an NLM request and while some code tests that, other code does not.

Rather than using rqstp->rq_client pass the client and gssclient explicitly to __fh_verify and then to nfsd_set_fh_dentry().

The final places where __fh_verify unconditionally dereferences rqstp involve checking if the connection is suitably secure. They look at rqstp->rq_xprt which is not meaningful in the target use case of "localio" NFS in which the client talks directly to the local server. So have these always succeed when rqstp is NULL.

Lastly, 4 associated tracepoints are only used if rqstp is not NULL (this is a stop-gap that should be properly fixed so localio also benefits from the utility these tracepoints provide when debugging fh_verify issues).

---
 fs/nfsd/export.c |   8 ++-
 fs/nfsd/nfsfh.c  | 124 ++++++++++++++++++++++++++++-------------------
 2 files changed, 82 insertions(+), 50 deletions(-)
fs/nfsd/export.c +7 -1
index 7bb4f2075ac5..fe36f441d1d9 100644
@@ -1077,7 +1077,13 @@ static struct svc_export *exp_find(struct cache_detail *cd,
 __be32 check_nfsd_access(struct svc_export *exp, struct svc_rqst *rqstp)
 {
 	struct exp_flavor_info *f, *end = exp->ex_flavors + exp->ex_nflavors;
-	struct svc_xprt *xprt = rqstp->rq_xprt;
+	struct svc_xprt *xprt;
+
+	if (!rqstp)
+		/* Always allow LOCALIO */
+		return 0;
+
+	xprt = rqstp->rq_xprt;
 
 	if (exp->ex_xprtsec_modes & NFSEXP_XPRTSEC_NONE) {
 		if (!test_bit(XPT_TLS_SESSION, &xprt->xpt_flags))
fs/nfsd/nfsfh.c +75 -49

218 lines not on this page · raw message

Signature · 1 line
2.44.0
#5 lore raw

Re: [PATCH v13 04/19] nfsd: factor out __fh_verify to allow NULL rqstp to be passed

Quoted 23 lines · Mike Snitzer

From: NeilBrown <[email protected]>

__fh_verify() offers an interface like fh_verify() but doesn't require a struct svc_rqst *, instead it also takes the specific parts as explicit required arguments. So it is safe to call __fh_verify() with a NULL rqstp, but the net, cred, and client args must not be NULL.

__fh_verify() does not use SVC_NET(), nor does the functions it calls.

Rather then depending on rqstp->rq_vers to determine nfs version, pass it in explicitly. This removes another dependency on rqstp and ensures the correct version is checked. The rqstp can be for an NLM request and while some code tests that, other code does not.

Rather than using rqstp->rq_client pass the client and gssclient explicitly to __fh_verify and then to nfsd_set_fh_dentry().

The final places where __fh_verify unconditionally dereferences rqstp involve checking if the connection is suitably secure. They look at rqstp->rq_xprt which is not meaningful in the target use case of "localio" NFS in which the client talks directly to the local server. So have these always succeed when rqstp is NULL.

Quoted 8 lines

Lastly, 4 associated tracepoints are only used if rqstp is not NULL (this is a stop-gap that should be properly fixed so localio also benefits from the utility these tracepoints provide when debugging fh_verify issues).

IMO this patch needs to be split up. There are several changes that need separate explanation and rationale, and the changes here need to be individually bisectable.

If you prefer, I can provide a patch series that replaces this one patch, or Neil could provide it if he wants.

A few more specific comments below.

Quoted 11 lines
---
 fs/nfsd/export.c |   8 ++-
 fs/nfsd/nfsfh.c  | 124 ++++++++++++++++++++++++++++-------------------
 2 files changed, 82 insertions(+), 50 deletions(-)
fs/nfsd/export.c +0 -0
index 7bb4f2075ac5..fe36f441d1d9 100644
@@ -1077,7 +1077,13 @@ static struct svc_export *exp_find(struct cache_detail *cd,
 __be32 check_nfsd_access(struct svc_export *exp, struct svc_rqst *rqstp)

fs/nfsd/export.c static struct svc_export *exp_find(struct cache_detail *cd, · line 1086

Quoted 10 lines
fs/nfsd/export.c +7 -1
 {
 	struct exp_flavor_info *f, *end = exp->ex_flavors + exp->ex_nflavors;
-	struct svc_xprt *xprt = rqstp->rq_xprt;
+	struct svc_xprt *xprt;
+
+	if (!rqstp)
+		/* Always allow LOCALIO */
+		return 0;
+
+	xprt = rqstp->rq_xprt;

check_nfsd_access() is a public API, so it needs a kdoc comment.

These changes should be split into a separate patch with a clear rationale of why “Always allow LOCALIO” is secure and appropriate to do.

Quoted 38 lines
fs/nfsd/export.c +0 -0
 	if (exp->ex_xprtsec_modes & NFSEXP_XPRTSEC_NONE) {
 		if (!test_bit(XPT_TLS_SESSION, &xprt->xpt_flags))
fs/nfsd/nfsfh.c +7 -6
index 50d23d56f403..19e173187ab9 100644
@@ -87,23 +87,24 @@ nfsd_mode_check(struct dentry *dentry, umode_t requested)
 	return nfserr_wrong_type;
 }
 
-static bool nfsd_originating_port_ok(struct svc_rqst *rqstp, int flags)
+static bool nfsd_originating_port_ok(struct svc_rqst *rqstp,
+				     struct svc_cred *cred,
+				     struct svc_export *exp)
 {
-	if (flags & NFSEXP_INSECURE_PORT)
+	if (nfsexp_flags(cred, exp) & NFSEXP_INSECURE_PORT)
 		return true;
 	/* We don't require gss requests to use low ports: */
-	if (rqstp->rq_cred.cr_flavor >= RPC_AUTH_GSS)
+	if (cred->cr_flavor >= RPC_AUTH_GSS)
 		return true;
 	return test_bit(RQ_SECURE, &rqstp->rq_flags);
 }
 
 static __be32 nfsd_setuser_and_check_port(struct svc_rqst *rqstp,
+					  struct svc_cred *cred,
 					  struct svc_export *exp)
 {
-	int flags = nfsexp_flags(&rqstp->rq_cred, exp);
-
 	/* Check if the request originated from a secure port. */
-	if (!nfsd_originating_port_ok(rqstp, flags)) {
+	if (rqstp && !nfsd_originating_port_ok(rqstp, cred, exp)) {
 		RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
 		dprintk("nfsd: request from insecure port %s!\n",
 		        svc_print_addr(rqstp, buf, sizeof(buf)));

fs/nfsd/nfsfh.c static __be32 nfsd_setuser_and_check_port(struct svc_rqst *rqstp, · line 116

Quoted 6 lines
fs/nfsd/nfsfh.c +1 -1
@@ -111,7 +112,7 @@ static __be32 nfsd_setuser_and_check_port(struct svc_rqst *rqstp,
 	}
 
 	/* Set user creds for this exportpoint */
-	return nfserrno(nfsd_setuser(&rqstp->rq_cred, exp));
+	return nfserrno(nfsd_setuser(cred, exp));
 }

Just for cleanliness, the above hunks could be in their own patch, labeled as a refactoring change.

Quoted 16 lines
fs/nfsd/nfsfh.c +5 -1
 static inline __be32 check_pseudo_root(struct dentry *dentry,
@@ -141,7 +142,11 @@ static inline __be32 check_pseudo_root(struct dentry *dentry,
  * dentry.  On success, the results are used to set fh_export and
  * fh_dentry.
  */
-static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp)
+static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct net *net,
+				 struct svc_cred *cred, int nfs_vers,
+				 struct auth_domain *client,
+				 struct auth_domain *gssclient,
+				 struct svc_fh *fhp)
 {
 	struct knfsd_fh	*fh = &fhp->fh_handle;
 	struct fid *fid = NULL;

fs/nfsd/nfsfh.c static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp) · line 193

Quoted 8 lines
fs/nfsd/nfsfh.c +2 -2
@@ -183,14 +188,15 @@ static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp)
 	data_left -= len;
 	if (data_left < 0)
 		return error;
-	exp = rqst_exp_find(&rqstp->rq_chandle, SVC_NET(rqstp),
-			    rqstp->rq_client, rqstp->rq_gssclient,
+	exp = rqst_exp_find(rqstp ? &rqstp->rq_chandle : NULL,
+			    net, client, gssclient,
 			    fh->fh_fsid_type, fh->fh_fsid);

Question: Would rqst_exp_find() be the function that would prevent a LOCALIO open to a file handle where the client's IP address is not listed on the export?

I don't really see how IP address-related export access control is being enforced, but it's possible I'm missing something.

Quoted 30 lines
fs/nfsd/nfsfh.c +5 -3
 	fid = (struct fid *)(fh->fh_fsid + len);
 
 	error = nfserr_stale;
 	if (IS_ERR(exp)) {
-		trace_nfsd_set_fh_dentry_badexport(rqstp, fhp, PTR_ERR(exp));
+		if (rqstp)
+			trace_nfsd_set_fh_dentry_badexport(rqstp, fhp, PTR_ERR(exp));
 
 		if (PTR_ERR(exp) == -ENOENT)
 			return error;
@@ -219,7 +225,7 @@ static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp)
 		put_cred(override_creds(new));
 		put_cred(new);
 	} else {
-		error = nfsd_setuser_and_check_port(rqstp, exp);
+		error = nfsd_setuser_and_check_port(rqstp, cred, exp);
 		if (error)
 			goto out;
 	}
@@ -238,7 +244,8 @@ static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp)
 						data_left, fileid_type, 0,
 						nfsd_acceptable, exp);
 		if (IS_ERR_OR_NULL(dentry)) {
-			trace_nfsd_set_fh_dentry_badhandle(rqstp, fhp,
+			if (rqstp)
+				trace_nfsd_set_fh_dentry_badhandle(rqstp, fhp,
 					dentry ?  PTR_ERR(dentry) : -ESTALE);
 			switch (PTR_ERR(dentry)) {
 			case -ENOMEM:

fs/nfsd/nfsfh.c static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp) · line 279

Quoted 8 lines
fs/nfsd/nfsfh.c +1 -1
@@ -266,7 +273,7 @@ static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp)
 	fhp->fh_dentry = dentry;
 	fhp->fh_export = exp;
 
-	switch (rqstp->rq_vers) {
+	switch (nfs_vers) {
 	case 4:
 		if (dentry->d_sb->s_export_op->flags & EXPORT_OP_NOATOMIC_ATTR)
 			fhp->fh_no_atomic_attr = true;

@nfs_vers is actually the new FH’s NFS version; that needs to be better documented because this is NFS version-specific code and I guess it needs to stay in here since it parametrizes the form of the new file handle.

Perhaps these cases need to be wrapped with ifdefs.

IMO all callers should supply this value, rather than manufacturing it in fh_verify(). Ie, "I want an NFSv3 file handle, please."

That should be a separate patch, it will be somewhat less than a surgical change.

Quoted 21 lines
fs/nfsd/nfsfh.c +0 -17
@@ -293,50 +300,29 @@ static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp)
 	return error;
 }
 
-/**
- * fh_verify - filehandle lookup and access checking
- * @rqstp: pointer to current rpc request
- * @fhp: filehandle to be verified
- * @type: expected type of object pointed to by filehandle
- * @access: type of access needed to object
- *
- * Look up a dentry from the on-the-wire filehandle, check the client's
- * access to the export, and set the current task's credentials.
- *
- * Regardless of success or failure of fh_verify(), fh_put() should be
- * called on @fhp when the caller is finished with the filehandle.
- *
- * fh_verify() may be called multiple times on a given filehandle, for
- * example, when processing an NFSv4 compound.  The first call will look
- * up a dentry using the on-the-wire filehandle.  Subsequent calls will
- * skip the lookup and just perform the other checks and possibly change

fs/nfsd/nfsfh.c static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp) · line 322

Quoted 10 lines
fs/nfsd/nfsfh.c +0 -10
- * the current task's credentials.
- *
- * @type specifies the type of object expected using one of the S_IF*
- * constants defined in include/linux/stat.h.  The caller may use zero
- * to indicate that it doesn't care, or a negative integer to indicate
- * that it expects something not of the given type.
- *
- * @access is formed from the NFSD_MAY_* constants defined in
- * fs/nfsd/vfs.h.
- */

See comment on 5/N: since that patch makes this a public API again, consider not removing this kdoc comment but rather updating it.

Quoted 100 lines
fs/nfsd/nfsfh.c +54 -8
-__be32
-fh_verify(struct svc_rqst *rqstp, struct svc_fh *fhp, umode_t type, int access)
+static __be32
+__fh_verify(struct svc_rqst *rqstp,
+	    struct net *net, struct svc_cred *cred,
+	    int nfs_vers, struct auth_domain *client,
+	    struct auth_domain *gssclient,
+	    struct svc_fh *fhp, umode_t type, int access)
 {
-	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
+	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 	struct svc_export *exp = NULL;
 	struct dentry	*dentry;
 	__be32		error;
 
 	if (!fhp->fh_dentry) {
-		error = nfsd_set_fh_dentry(rqstp, fhp);
+		error = nfsd_set_fh_dentry(rqstp, net, cred, nfs_vers,
+					   client, gssclient, fhp);
 		if (error)
 			goto out;
 	}
 	dentry = fhp->fh_dentry;
 	exp = fhp->fh_export;
 
-	trace_nfsd_fh_verify(rqstp, fhp, type, access);
+	if (rqstp)
+		trace_nfsd_fh_verify(rqstp, fhp, type, access);
 
 	/*
 	 * We still have to do all these permission checks, even when
@@ -358,7 +344,7 @@ fh_verify(struct svc_rqst *rqstp, struct svc_fh *fhp, umode_t type, int access)
 	if (error)
 		goto out;
 
-	error = nfsd_setuser_and_check_port(rqstp, exp);
+	error = nfsd_setuser_and_check_port(rqstp, cred, exp);
 	if (error)
 		goto out;
 
@@ -388,14 +374,54 @@ fh_verify(struct svc_rqst *rqstp, struct svc_fh *fhp, umode_t type, int access)
 
 skip_pseudoflavor_check:
 	/* Finally, check access permissions. */
-	error = nfsd_permission(&rqstp->rq_cred, exp, dentry, access);
+	error = nfsd_permission(cred, exp, dentry, access);
 out:
-	trace_nfsd_fh_verify_err(rqstp, fhp, type, access, error);
+	if (rqstp)
+		trace_nfsd_fh_verify_err(rqstp, fhp, type, access, error);
 	if (error == nfserr_stale)
 		nfsd_stats_fh_stale_inc(nn, exp);
 	return error;
 }
 
+/**
+ * fh_verify - filehandle lookup and access checking
+ * @rqstp: pointer to current rpc request
+ * @fhp: filehandle to be verified
+ * @type: expected type of object pointed to by filehandle
+ * @access: type of access needed to object
+ *
+ * Look up a dentry from the on-the-wire filehandle, check the client's
+ * access to the export, and set the current task's credentials.
+ *
+ * Regardless of success or failure of fh_verify(), fh_put() should be
+ * called on @fhp when the caller is finished with the filehandle.
+ *
+ * fh_verify() may be called multiple times on a given filehandle, for
+ * example, when processing an NFSv4 compound.  The first call will look
+ * up a dentry using the on-the-wire filehandle.  Subsequent calls will
+ * skip the lookup and just perform the other checks and possibly change
+ * the current task's credentials.
+ *
+ * @type specifies the type of object expected using one of the S_IF*
+ * constants defined in include/linux/stat.h.  The caller may use zero
+ * to indicate that it doesn't care, or a negative integer to indicate
+ * that it expects something not of the given type.
+ *
+ * @access is formed from the NFSD_MAY_* constants defined in
+ * fs/nfsd/vfs.h.
+ */
+__be32
+fh_verify(struct svc_rqst *rqstp, struct svc_fh *fhp, umode_t type, int access)
+{
+	int nfs_vers;
+	if (rqstp->rq_prog == NFS_PROGRAM)
+		nfs_vers = rqstp->rq_vers;
+	else /* must be NLM */
+		nfs_vers = rqstp->rq_vers == 4 ? 3 : 2;
+	return __fh_verify(rqstp, SVC_NET(rqstp), &rqstp->rq_cred, nfs_vers,
+			   rqstp->rq_client, rqstp->rq_gssclient,
+			   fhp, type, access);
+}
 
 /*
  * Compose a file handle for an NFS reply.
Signature · 1 line
2.44.0
Signature · 1 line
Chuck Lever
#6 lore raw

Re: [PATCH v13 04/19] nfsd: factor out __fh_verify to allow NULL rqstp to be passed

Quoted 8 lines · Chuck Lever

fs/nfsd/nfsfh.c static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp) · line 193

Quoted 3 lines · Mike Snitzer
fs/nfsd/nfsfh.c +2 -0
+	exp = rqst_exp_find(rqstp ? &rqstp->rq_chandle : NULL,
+			    net, client, gssclient,
 			    fh->fh_fsid_type, fh->fh_fsid);

Question: Would rqst_exp_find() be the function that would prevent a LOCALIO open to a file handle where the client's IP address is not listed on the export?

Yes.

Quoted 3 lines

I don't really see how IP address-related export access control is being enforced, but it's possible I'm missing something.

The "client" is key. The LOCALIO RPC protocol allows the server to determine a "client" which matches the network connection. It passes this to the client code which uses it for future authentication.

Quoted 2 lines

See comment on 5/N: since that patch makes this a public API again, consider not removing this kdoc comment but rather updating it.

What exactly do you consider to be a "public API"?? Anything without "static"? That seems somewhat arbitrary.

I think of __fh_verify() as a private API used by fh_verify() and nfsd_file_acquire_local() and nothing else.

It seems pointless duplication the documentation for __fh_verify() and fh_verify(). Maybe one could refer to the other "fh_verify is like fh_verify except ....."

??

Quoted 6 lines

fs/nfsd/nfsfh.c static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp) · line 304

Quoted 6 lines
fs/nfsd/nfsfh.c +2 -2
-__be32
-fh_verify(struct svc_rqst *rqstp, struct svc_fh *fhp, umode_t type, int access)
+static __be32
+__fh_verify(struct svc_rqst *rqstp,

Thanks,
NeilBrown

#7 lore raw

Re: [PATCH v13 04/19] nfsd: factor out __fh_verify to allow NULL rqstp to be passed

Quoted 6 lines · NeilBrown
Quoted 2 lines · Chuck Lever

See comment on 5/N: since that patch makes this a public API again, consider not removing this kdoc comment but rather updating it.

What exactly do you consider to be a "public API"?? Anything without "static"? That seems somewhat arbitrary.

Anything that is invoked from another, distinct, area of code presents an API boundary.

Quoted 2 lines

I think of __fh_verify() as a private API used by fh_verify() and nfsd_file_acquire_local() and nothing else.

That is indeed the current situation. But practically speaking, there's nothing that C can do to prevent other call sites from appearing over time.

Quoted 2 lines

It seems pointless duplication the documentation for __fh_verify() and fh_verify().

We have that duplication already for the family of nfsd_file_acquire APIs, for example.

These two, however, take somewhat different parameters, and somewhere (in the code) you have to document things like "this one expects a non-NULL @rqstp". Some API documentation, even for "private" APIs, is valuable.

Quoted 2 lines

Maybe one could refer to the other "fh_verify is like fh_verify except ....."

Highlighting the differences is certainly helpful for human readers.

Perhaps it doesn't have to be a kdoc-style comment, but that's generally the kind of comment a developer would expect to explain how and when to invoke a function.

Signature · 1 line
Chuck Lever
#8 lore raw

Re: [PATCH v13 04/19] nfsd: factor out __fh_verify to allow NULL rqstp to be passed

In reply to NeilBrown
Quoted 14 lines · NeilBrown
Quoted 2 lines · Chuck Lever

See comment on 5/N: since that patch makes this a public API again, consider not removing this kdoc comment but rather updating it.

What exactly do you consider to be a "public API"?? Anything without "static"? That seems somewhat arbitrary.

I think of __fh_verify() as a private API used by fh_verify() and nfsd_file_acquire_local() and nothing else.

It seems pointless duplication the documentation for __fh_verify() and fh_verify(). Maybe one could refer to the other "fh_verify is like fh_verify except ....."

Quoted 8 lines

??

Quoted 6 lines

fs/nfsd/nfsfh.c static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp) · line 304

Quoted 6 lines
fs/nfsd/nfsfh.c +2 -2
-__be32
-fh_verify(struct svc_rqst *rqstp, struct svc_fh *fhp, umode_t type, int access)
+static __be32
+__fh_verify(struct svc_rqst *rqstp,

An alternative would be to leave __fh_verify() as a static, and then add an fh_verify_local() API, echoing nfsd_file_acquire_local(), and then give that API a kdoc comment.

That would make it clear who the intended consumer is.

Signature · 1 line
Chuck Lever
#9 lore raw

Re: [PATCH v13 04/19] nfsd: factor out __fh_verify to allow NULL rqstp to be passed

Quoted 28 lines · Chuck Lever
Quoted 14 lines · NeilBrown
Quoted 2 lines · Chuck Lever

See comment on 5/N: since that patch makes this a public API again, consider not removing this kdoc comment but rather updating it.

What exactly do you consider to be a "public API"?? Anything without "static"? That seems somewhat arbitrary.

I think of __fh_verify() as a private API used by fh_verify() and nfsd_file_acquire_local() and nothing else.

It seems pointless duplication the documentation for __fh_verify() and fh_verify(). Maybe one could refer to the other "fh_verify is like fh_verify except ....."

Quoted 8 lines

??

Quoted 6 lines

fs/nfsd/nfsfh.c static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp) · line 304

Quoted 6 lines
fs/nfsd/nfsfh.c +2 -2
-__be32
-fh_verify(struct svc_rqst *rqstp, struct svc_fh *fhp, umode_t type, int access)
+static __be32
+__fh_verify(struct svc_rqst *rqstp,

An alternative would be to leave __fh_verify() as a static, and then add an fh_verify_local() API, echoing nfsd_file_acquire_local(), and then give that API a kdoc comment.

Quoted 1 line

That would make it clear who the intended consumer is.

I ran with this idea, definite improvement, you'll find the changes folded in to the relevant patch (with the same subject) in v14.

Thanks,
Mike

#10 lore raw

Re: [PATCH v13 04/19] nfsd: factor out __fh_verify to allow NULL rqstp to be passed

Quoted 29 lines · Mike Snitzer
Quoted 28 lines · Chuck Lever
Quoted 14 lines · NeilBrown
Quoted 2 lines · Chuck Lever

See comment on 5/N: since that patch makes this a public API again, consider not removing this kdoc comment but rather updating it.

What exactly do you consider to be a "public API"?? Anything without "static"? That seems somewhat arbitrary.

I think of __fh_verify() as a private API used by fh_verify() and nfsd_file_acquire_local() and nothing else.

It seems pointless duplication the documentation for __fh_verify() and fh_verify(). Maybe one could refer to the other "fh_verify is like fh_verify except ....."

Quoted 8 lines

??

Quoted 6 lines

fs/nfsd/nfsfh.c static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp) · line 304

Quoted 6 lines
fs/nfsd/nfsfh.c +2 -2
-__be32
-fh_verify(struct svc_rqst *rqstp, struct svc_fh *fhp, umode_t type, int access)
+static __be32
+__fh_verify(struct svc_rqst *rqstp,

An alternative would be to leave __fh_verify() as a static, and then add an fh_verify_local() API, echoing nfsd_file_acquire_local(), and then give that API a kdoc comment.

Quoted 4 lines
Quoted 1 line

That would make it clear who the intended consumer is.

I ran with this idea, definite improvement, you'll find the changes folded in to the relevant patch (with the same subject) in v14.

Err, sorry, in v14 it'll be patch 9 with subject:
"nfsd: add nfsd_file_acquire_local()"

#11 lore raw

Re: [PATCH v13 04/19] nfsd: factor out __fh_verify to allow NULL rqstp to be passed

In reply to Chuck Lever
Quoted 33 lines · Chuck Lever
Quoted 23 lines · Mike Snitzer

From: NeilBrown <[email protected]>

__fh_verify() offers an interface like fh_verify() but doesn't require a struct svc_rqst *, instead it also takes the specific parts as explicit required arguments. So it is safe to call __fh_verify() with a NULL rqstp, but the net, cred, and client args must not be NULL.

__fh_verify() does not use SVC_NET(), nor does the functions it calls.

Rather then depending on rqstp->rq_vers to determine nfs version, pass it in explicitly. This removes another dependency on rqstp and ensures the correct version is checked. The rqstp can be for an NLM request and while some code tests that, other code does not.

Rather than using rqstp->rq_client pass the client and gssclient explicitly to __fh_verify and then to nfsd_set_fh_dentry().

The final places where __fh_verify unconditionally dereferences rqstp involve checking if the connection is suitably secure. They look at rqstp->rq_xprt which is not meaningful in the target use case of "localio" NFS in which the client talks directly to the local server. So have these always succeed when rqstp is NULL.

Quoted 8 lines

Lastly, 4 associated tracepoints are only used if rqstp is not NULL (this is a stop-gap that should be properly fixed so localio also benefits from the utility these tracepoints provide when debugging fh_verify issues).

Quoted 6 lines

IMO this patch needs to be split up. There are several changes that need separate explanation and rationale, and the changes here need to be individually bisectable.

If you prefer, I can provide a patch series that replaces this one patch, or Neil could provide it if he wants.

I'm probably not the best person to take a crack at splitting this patch up.

Neil initially had factored out N patches, but they weren't fully baked so when I had to fix the code up it became a challenge to sprinkle fixes across the N patches. Because they were all pretty interdependent.

In the end, all these changes are in service to allowing for the possibility that the rqstp not available (NULL). So it made sense to fold them together.

I really don't see how factoring these changes out to N patches makes them useful for bisection (you need all of them to test the case when rqstp is NULL, and a partial application of changes to track down a rqstp-full regression really isn't such a concern given fh_verify() fully passes all args to __fh_verify so status-quo preserved).

All said, if your intuition and experience makes you feel splitting this patch up is needed then I'm fine with it and I welcome your or Neil's contribtion. It is fiddley work though, so having had my own challenges with the code when these changes were split out makes me hesitant to jump on splitting them out again.

Hope I've explained myself clearly... not being confrontational, dismissive or anything else. ;)

Quoted 27 lines

A few more specific comments below.

Quoted 11 lines
---
 fs/nfsd/export.c |   8 ++-
 fs/nfsd/nfsfh.c  | 124 ++++++++++++++++++++++++++++-------------------
 2 files changed, 82 insertions(+), 50 deletions(-)
fs/nfsd/export.c +0 -0
index 7bb4f2075ac5..fe36f441d1d9 100644
@@ -1077,7 +1077,13 @@ static struct svc_export *exp_find(struct cache_detail *cd,
 __be32 check_nfsd_access(struct svc_export *exp, struct svc_rqst *rqstp)

fs/nfsd/export.c static struct svc_export *exp_find(struct cache_detail *cd, · line 1086

Quoted 10 lines
fs/nfsd/export.c +7 -1
 {
 	struct exp_flavor_info *f, *end = exp->ex_flavors + exp->ex_nflavors;
-	struct svc_xprt *xprt = rqstp->rq_xprt;
+	struct svc_xprt *xprt;
+
+	if (!rqstp)
+		/* Always allow LOCALIO */
+		return 0;
+
+	xprt = rqstp->rq_xprt;

check_nfsd_access() is a public API, so it needs a kdoc comment.

Quoted 3 lines

These changes should be split into a separate patch with a clear rationale of why "Always allow LOCALIO" is secure and appropriate to do.

Separate patch aside, I'll try to improve that comment.

Thanks,
Mike

#12 lore raw

Re: [PATCH v13 04/19] nfsd: factor out __fh_verify to allow NULL rqstp to be passed

Quoted 59 lines · Mike Snitzer
Quoted 33 lines · Chuck Lever
Quoted 23 lines · Mike Snitzer

From: NeilBrown <[email protected]>

__fh_verify() offers an interface like fh_verify() but doesn't require a struct svc_rqst *, instead it also takes the specific parts as explicit required arguments. So it is safe to call __fh_verify() with a NULL rqstp, but the net, cred, and client args must not be NULL.

__fh_verify() does not use SVC_NET(), nor does the functions it calls.

Rather then depending on rqstp->rq_vers to determine nfs version, pass it in explicitly. This removes another dependency on rqstp and ensures the correct version is checked. The rqstp can be for an NLM request and while some code tests that, other code does not.

Rather than using rqstp->rq_client pass the client and gssclient explicitly to __fh_verify and then to nfsd_set_fh_dentry().

The final places where __fh_verify unconditionally dereferences rqstp involve checking if the connection is suitably secure. They look at rqstp->rq_xprt which is not meaningful in the target use case of "localio" NFS in which the client talks directly to the local server. So have these always succeed when rqstp is NULL.

Quoted 8 lines

Lastly, 4 associated tracepoints are only used if rqstp is not NULL (this is a stop-gap that should be properly fixed so localio also benefits from the utility these tracepoints provide when debugging fh_verify issues).

Quoted 6 lines

IMO this patch needs to be split up. There are several changes that need separate explanation and rationale, and the changes here need to be individually bisectable.

If you prefer, I can provide a patch series that replaces this one patch, or Neil could provide it if he wants.

I'm probably not the best person to take a crack at splitting this patch up.

Neil initially had factored out N patches, but they weren't fully baked so when I had to fix the code up it became a challenge to sprinkle fixes across the N patches. Because they were all pretty interdependent.

In the end, all these changes are in service to allowing for the possibility that the rqstp not available (NULL). So it made sense to fold them together.

I really don't see how factoring these changes out to N patches makes them useful for bisection (you need all of them to test the case when rqstp is NULL, and a partial application of changes to track down a rqstp-full regression really isn't such a concern given fh_verify() fully passes all args to __fh_verify so status-quo preserved).

Quoted 8 lines

All said, if your intuition and experience makes you feel splitting this patch up is needed then I'm fine with it and I welcome your or Neil's contribtion. It is fiddley work though, so having had my own challenges with the code when these changes were split out makes me hesitant to jump on splitting them out again.

Hope I've explained myself clearly... not being confrontational, dismissive or anything else. ;)

True, this isn't an enormous single patch, but you gotta draw that line somewhere.

The goal is to make these changes as small and atomic as possible so it becomes easy to spot a mistake or bisect to find unintended behavior introduced in the non-LOCALIO case. This is a code path that is heavily utilized by NFSD so it pays to take some defensive precautions.

Generally I find that a typo or hidden assumption magically appears when I've engaged in this kind of refactoring exercise. I've got a good toolchain that should make this quick work.

Quoted 17 lines
Quoted 17 lines
Quoted 4 lines
fs/nfsd/export.c +0 -0
@@ -1077,7 +1077,13 @@ static struct svc_export *exp_find(struct cache_detail *cd,
 __be32 check_nfsd_access(struct svc_export *exp, struct svc_rqst *rqstp)

fs/nfsd/export.c static struct svc_export *exp_find(struct cache_detail *cd, · line 1086

Quoted 10 lines
fs/nfsd/export.c +7 -1
 {
 	struct exp_flavor_info *f, *end = exp->ex_flavors + exp->ex_nflavors;
-	struct svc_xprt *xprt = rqstp->rq_xprt;
+	struct svc_xprt *xprt;
+
+	if (!rqstp)
+		/* Always allow LOCALIO */
+		return 0;
+
+	xprt = rqstp->rq_xprt;

check_nfsd_access() is a public API, so it needs a kdoc comment.

Quoted 5 lines
Quoted 3 lines

These changes should be split into a separate patch with a clear rationale of why "Always allow LOCALIO" is secure and appropriate to do.

Separate patch aside, I'll try to improve that comment.

Post the comment update here and I can work that in.

Signature · 1 line
Chuck Lever
#13 lore raw

Re: [PATCH v13 04/19] nfsd: factor out __fh_verify to allow NULL rqstp to be passed

Quoted 78 lines · Chuck Lever
Quoted 59 lines · Mike Snitzer
Quoted 33 lines · Chuck Lever
Quoted 23 lines · Mike Snitzer

From: NeilBrown <[email protected]>

__fh_verify() offers an interface like fh_verify() but doesn't require a struct svc_rqst *, instead it also takes the specific parts as explicit required arguments. So it is safe to call __fh_verify() with a NULL rqstp, but the net, cred, and client args must not be NULL.

__fh_verify() does not use SVC_NET(), nor does the functions it calls.

Rather then depending on rqstp->rq_vers to determine nfs version, pass it in explicitly. This removes another dependency on rqstp and ensures the correct version is checked. The rqstp can be for an NLM request and while some code tests that, other code does not.

Rather than using rqstp->rq_client pass the client and gssclient explicitly to __fh_verify and then to nfsd_set_fh_dentry().

The final places where __fh_verify unconditionally dereferences rqstp involve checking if the connection is suitably secure. They look at rqstp->rq_xprt which is not meaningful in the target use case of "localio" NFS in which the client talks directly to the local server. So have these always succeed when rqstp is NULL.

Quoted 8 lines

Lastly, 4 associated tracepoints are only used if rqstp is not NULL (this is a stop-gap that should be properly fixed so localio also benefits from the utility these tracepoints provide when debugging fh_verify issues).

Quoted 6 lines

IMO this patch needs to be split up. There are several changes that need separate explanation and rationale, and the changes here need to be individually bisectable.

If you prefer, I can provide a patch series that replaces this one patch, or Neil could provide it if he wants.

I'm probably not the best person to take a crack at splitting this patch up.

Neil initially had factored out N patches, but they weren't fully baked so when I had to fix the code up it became a challenge to sprinkle fixes across the N patches. Because they were all pretty interdependent.

In the end, all these changes are in service to allowing for the possibility that the rqstp not available (NULL). So it made sense to fold them together.

I really don't see how factoring these changes out to N patches makes them useful for bisection (you need all of them to test the case when rqstp is NULL, and a partial application of changes to track down a rqstp-full regression really isn't such a concern given fh_verify() fully passes all args to __fh_verify so status-quo preserved).

Quoted 8 lines

All said, if your intuition and experience makes you feel splitting this patch up is needed then I'm fine with it and I welcome your or Neil's contribtion. It is fiddley work though, so having had my own challenges with the code when these changes were split out makes me hesitant to jump on splitting them out again.

Hope I've explained myself clearly... not being confrontational, dismissive or anything else. ;)

True, this isn't an enormous single patch, but you gotta draw that line somewhere.

The goal is to make these changes as small and atomic as possible so it becomes easy to spot a mistake or bisect to find unintended behavior introduced in the non-LOCALIO case. This is a code path that is heavily utilized by NFSD so it pays to take some defensive precautions.

Quoted 3 lines

Generally I find that a typo or hidden assumption magically appears when I've engaged in this kind of refactoring exercise. I've got a good toolchain that should make this quick work.

Awesome.

Quoted 17 lines
Quoted 17 lines
Quoted 17 lines
Quoted 4 lines
fs/nfsd/export.c +0 -0
@@ -1077,7 +1077,13 @@ static struct svc_export *exp_find(struct cache_detail *cd,
 __be32 check_nfsd_access(struct svc_export *exp, struct svc_rqst *rqstp)

fs/nfsd/export.c static struct svc_export *exp_find(struct cache_detail *cd, · line 1086

Quoted 10 lines
fs/nfsd/export.c +7 -1
 {
 	struct exp_flavor_info *f, *end = exp->ex_flavors + exp->ex_nflavors;
-	struct svc_xprt *xprt = rqstp->rq_xprt;
+	struct svc_xprt *xprt;
+
+	if (!rqstp)
+		/* Always allow LOCALIO */
+		return 0;
+
+	xprt = rqstp->rq_xprt;

check_nfsd_access() is a public API, so it needs a kdoc comment.

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines

These changes should be split into a separate patch with a clear rationale of why "Always allow LOCALIO" is secure and appropriate to do.

Separate patch aside, I'll try to improve that comment.

Post the comment update here and I can work that in.

Here is the incremental patch:

fs/nfsd/export.c +14 -2
index fe36f441d1d9..8b54f66f0e0f 100644
@@ -1074,14 +1074,26 @@ static struct svc_export *exp_find(struct cache_detail *cd,
 	return exp;
 }
 
+/**
+ * check_nfsd_access - check if access to export is allowed.
+ * @exp: svc_export that is being accessed.
+ * @rqstp: svc_rqst attempting to access @exp (will be NULL for LOCALIO).
+ *
+ * Returns 0 if access is granted or nfserr_wrongsec if access is denied.
+ */
 __be32 check_nfsd_access(struct svc_export *exp, struct svc_rqst *rqstp)
 {
 	struct exp_flavor_info *f, *end = exp->ex_flavors + exp->ex_nflavors;
 	struct svc_xprt *xprt;
 
-	if (!rqstp)
-		/* Always allow LOCALIO */
+	if (!rqstp) {
+		/*
+		 * The target use case for rqstp being NULL is LOCALIO,
+		 * which only supports AUTH_UNIX, so always allow LOCALIO
+		 * because the other checks below aren't applicable.
+		 */
 		return 0;
+	}
 
 	xprt = rqstp->rq_xprt;
 
#14 lore raw

PATCH v13 5/19 nfsd: add nfsd_file_acquire_local()

In reply to Mike Snitzer

From: NeilBrown <[email protected]>

nfsd_file_acquire_local() can be used to look up a file by filehandle without having a struct svc_rqst. This can be used by NFS LOCALIO to allow the NFS client to bypass the NFS protocol to directly access a file provided by the NFS server which is running in the same kernel.

In nfsd_file_do_acquire() care is taken to always use fh_verify() if rqstp is not NULL (as is the case for non-LOCALIO callers). Otherwise the non-LOCALIO callers will not supply the correct and required arguments to __fh_verify (e.g. nfs_vers is 0, gssclient isn't passed).

Also, use GC for nfsd_file returned by nfsd_file_acquire_local. GC offers performance improvements if/when a file is reopened before launderette cleans it from the filecache's LRU.

---
 fs/nfsd/filecache.c | 63 ++++++++++++++++++++++++++++++++++++++++-----
 fs/nfsd/filecache.h |  4 +++
 fs/nfsd/nfsfh.c     |  2 +-
 fs/nfsd/nfsfh.h     |  5 ++++
 4 files changed, 66 insertions(+), 8 deletions(-)
fs/nfsd/filecache.c +56 -7
index 9e9d246f993c..94ecb9ed0ed1 100644
@@ -982,12 +982,14 @@ nfsd_file_is_cached(struct inode *inode)
 }
 
 static __be32
-nfsd_file_do_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp,
+nfsd_file_do_acquire(struct svc_rqst *rqstp, struct net *net,
+		     struct svc_cred *cred, int nfs_vers,
+		     struct auth_domain *client,
+		     struct svc_fh *fhp,
 		     unsigned int may_flags, struct file *file,
 		     struct nfsd_file **pnf, bool want_gc)
 {
 	unsigned char need = may_flags & NFSD_FILE_MAY_MASK;
-	struct net *net = SVC_NET(rqstp);
 	struct nfsd_file *new, *nf;
 	bool stale_retry = true;
 	bool open_retry = true;
@@ -996,8 +998,13 @@ nfsd_file_do_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp,
 	int ret;
 
 retry:
-	status = fh_verify(rqstp, fhp, S_IFREG,
-				may_flags|NFSD_MAY_OWNER_OVERRIDE);
+	if (rqstp) {
+		status = fh_verify(rqstp, fhp, S_IFREG,
+				   may_flags|NFSD_MAY_OWNER_OVERRIDE);
+	} else {
+		status = __fh_verify(NULL, net, cred, nfs_vers, client, NULL, fhp,
+				     S_IFREG, may_flags|NFSD_MAY_OWNER_OVERRIDE);
+	}
 	if (status != nfs_ok)
 		return status;
 	inode = d_inode(fhp->fh_dentry);
@@ -1143,7 +1150,8 @@ __be32
 nfsd_file_acquire_gc(struct svc_rqst *rqstp, struct svc_fh *fhp,
 		     unsigned int may_flags, struct nfsd_file **pnf)
 {
-	return nfsd_file_do_acquire(rqstp, fhp, may_flags, NULL, pnf, true);
+	return nfsd_file_do_acquire(rqstp, SVC_NET(rqstp), NULL, 0, NULL,
+				    fhp, may_flags, NULL, pnf, true);
 }
 
 /**
@@ -1167,7 +1175,47 @@ __be32
 nfsd_file_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp,
 		  unsigned int may_flags, struct nfsd_file **pnf)
 {
-	return nfsd_file_do_acquire(rqstp, fhp, may_flags, NULL, pnf, false);
+	return nfsd_file_do_acquire(rqstp, SVC_NET(rqstp), NULL, 0, NULL,
+				    fhp, may_flags, NULL, pnf, false);
+}
+
+/**
+ * nfsd_file_acquire_local - Get a struct nfsd_file with an open file for localio
+ * @net: The network namespace in which to perform a lookup
+ * @cred: the user credential with which to validate access
+ * @nfs_vers: NFS version number to assume for request
+ * @client: the auth_domain for LOCALIO lookup
+ * @fhp: the NFS filehandle of the file to be opened
+ * @may_flags: NFSD_MAY_ settings for the file
+ * @pnf: OUT: new or found "struct nfsd_file" object
+ *
+ * This file lookup interface provide access to a file given the
+ * filehandle and credential.  No connection-based authorisation
+ * is performed and in that way it is quite different to other
+ * file access mediated by nfsd.  It allows a kernel module such as the NFS
+ * client to reach across network and filesystem namespaces to access
+ * a file.  The security implications of this should be carefully
+ * considered before use.
+ *
+ * The nfsd_file object returned by this API is reference-counted
+ * and garbage-collected. The object is retained for a few
+ * seconds after the final nfsd_file_put() in case the caller
+ * wants to re-use it.
+ *
+ * Return values:
+ *   %nfs_ok - @pnf points to an nfsd_file with its reference
+ *   count boosted.
+ *
+ * On error, an nfsstat value in network byte order is returned.
+ */
+__be32
+nfsd_file_acquire_local(struct net *net, struct svc_cred *cred,
+			int nfs_vers, struct auth_domain *client,
+			struct svc_fh *fhp,
+			unsigned int may_flags, struct nfsd_file **pnf)
+{
+	return nfsd_file_do_acquire(NULL, net, cred, nfs_vers, client,
+				    fhp, may_flags, NULL, pnf, true);
 }
 
 /**
@@ -1193,7 +1241,8 @@ nfsd_file_acquire_opened(struct svc_rqst *rqstp, struct svc_fh *fhp,
 			 unsigned int may_flags, struct file *file,
 			 struct nfsd_file **pnf)
 {
-	return nfsd_file_do_acquire(rqstp, fhp, may_flags, file, pnf, false);
+	return nfsd_file_do_acquire(rqstp, SVC_NET(rqstp), NULL, 0, NULL,
+				    fhp, may_flags, file, pnf, false);
 }
 
 /*
fs/nfsd/filecache.h +4 -0
index 3fbec24eea6c..6dab41f8541e 100644
@@ -66,5 +66,9 @@ __be32 nfsd_file_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp,
 __be32 nfsd_file_acquire_opened(struct svc_rqst *rqstp, struct svc_fh *fhp,
 		  unsigned int may_flags, struct file *file,
 		  struct nfsd_file **nfp);
+__be32 nfsd_file_acquire_local(struct net *net, struct svc_cred *cred,
+			       int nfs_vers, struct auth_domain *client,
+			       struct svc_fh *fhp,
+			       unsigned int may_flags, struct nfsd_file **pnf);
 int nfsd_file_cache_stats_show(struct seq_file *m, void *v);
 #endif /* _FS_NFSD_FILECACHE_H */
fs/nfsd/nfsfh.c +1 -1
index 19e173187ab9..3635c0390cab 100644
@@ -300,7 +300,7 @@ static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct net *net,
 	return error;
 }
 
-static __be32
+__be32
 __fh_verify(struct svc_rqst *rqstp,
 	    struct net *net, struct svc_cred *cred,
 	    int nfs_vers, struct auth_domain *client,
fs/nfsd/nfsfh.h +5 -0
index 8d46e203d139..1429bee0ac1c 100644
@@ -217,6 +217,11 @@ extern char * SVCFH_fmt(struct svc_fh *fhp);
  * Function prototypes
  */
 __be32	fh_verify(struct svc_rqst *, struct svc_fh *, umode_t, int);
+__be32	__fh_verify(struct svc_rqst *rqstp,
+		    struct net *net, struct svc_cred *cred,
+		    int nfs_vers, struct auth_domain *client,
+		    struct auth_domain *gssclient,
+		    struct svc_fh *fhp, umode_t type, int access);
 __be32	fh_compose(struct svc_fh *, struct svc_export *, struct dentry *, struct svc_fh *);
 __be32	fh_update(struct svc_fh *);
 void	fh_put(struct svc_fh *);
Signature · 1 line
2.44.0
#15 lore raw

Re: [PATCH v13 05/19] nfsd: add nfsd_file_acquire_local()

Quoted 152 lines · Mike Snitzer

From: NeilBrown <[email protected]>

nfsd_file_acquire_local() can be used to look up a file by filehandle without having a struct svc_rqst. This can be used by NFS LOCALIO to allow the NFS client to bypass the NFS protocol to directly access a file provided by the NFS server which is running in the same kernel.

In nfsd_file_do_acquire() care is taken to always use fh_verify() if rqstp is not NULL (as is the case for non-LOCALIO callers). Otherwise the non-LOCALIO callers will not supply the correct and required arguments to __fh_verify (e.g. nfs_vers is 0, gssclient isn't passed).

Also, use GC for nfsd_file returned by nfsd_file_acquire_local. GC offers performance improvements if/when a file is reopened before launderette cleans it from the filecache's LRU.

---
 fs/nfsd/filecache.c | 63 ++++++++++++++++++++++++++++++++++++++++-----
 fs/nfsd/filecache.h |  4 +++
 fs/nfsd/nfsfh.c     |  2 +-
 fs/nfsd/nfsfh.h     |  5 ++++
 4 files changed, 66 insertions(+), 8 deletions(-)
fs/nfsd/filecache.c +56 -7
index 9e9d246f993c..94ecb9ed0ed1 100644
@@ -982,12 +982,14 @@ nfsd_file_is_cached(struct inode *inode)
 }
 
 static __be32
-nfsd_file_do_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp,
+nfsd_file_do_acquire(struct svc_rqst *rqstp, struct net *net,
+		     struct svc_cred *cred, int nfs_vers,
+		     struct auth_domain *client,
+		     struct svc_fh *fhp,
 		     unsigned int may_flags, struct file *file,
 		     struct nfsd_file **pnf, bool want_gc)
 {
 	unsigned char need = may_flags & NFSD_FILE_MAY_MASK;
-	struct net *net = SVC_NET(rqstp);
 	struct nfsd_file *new, *nf;
 	bool stale_retry = true;
 	bool open_retry = true;
@@ -996,8 +998,13 @@ nfsd_file_do_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp,
 	int ret;
 
 retry:
-	status = fh_verify(rqstp, fhp, S_IFREG,
-				may_flags|NFSD_MAY_OWNER_OVERRIDE);
+	if (rqstp) {
+		status = fh_verify(rqstp, fhp, S_IFREG,
+				   may_flags|NFSD_MAY_OWNER_OVERRIDE);
+	} else {
+		status = __fh_verify(NULL, net, cred, nfs_vers, client, NULL, fhp,
+				     S_IFREG, may_flags|NFSD_MAY_OWNER_OVERRIDE);
+	}
 	if (status != nfs_ok)
 		return status;
 	inode = d_inode(fhp->fh_dentry);
@@ -1143,7 +1150,8 @@ __be32
 nfsd_file_acquire_gc(struct svc_rqst *rqstp, struct svc_fh *fhp,
 		     unsigned int may_flags, struct nfsd_file **pnf)
 {
-	return nfsd_file_do_acquire(rqstp, fhp, may_flags, NULL, pnf, true);
+	return nfsd_file_do_acquire(rqstp, SVC_NET(rqstp), NULL, 0, NULL,
+				    fhp, may_flags, NULL, pnf, true);
 }
 
 /**
@@ -1167,7 +1175,47 @@ __be32
 nfsd_file_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp,
 		  unsigned int may_flags, struct nfsd_file **pnf)
 {
-	return nfsd_file_do_acquire(rqstp, fhp, may_flags, NULL, pnf, false);
+	return nfsd_file_do_acquire(rqstp, SVC_NET(rqstp), NULL, 0, NULL,
+				    fhp, may_flags, NULL, pnf, false);
+}
+
+/**
+ * nfsd_file_acquire_local - Get a struct nfsd_file with an open file for localio
+ * @net: The network namespace in which to perform a lookup
+ * @cred: the user credential with which to validate access
+ * @nfs_vers: NFS version number to assume for request
+ * @client: the auth_domain for LOCALIO lookup
+ * @fhp: the NFS filehandle of the file to be opened
+ * @may_flags: NFSD_MAY_ settings for the file
+ * @pnf: OUT: new or found "struct nfsd_file" object
+ *
+ * This file lookup interface provide access to a file given the
+ * filehandle and credential.  No connection-based authorisation
+ * is performed and in that way it is quite different to other
+ * file access mediated by nfsd.  It allows a kernel module such as the NFS
+ * client to reach across network and filesystem namespaces to access
+ * a file.  The security implications of this should be carefully
+ * considered before use.
+ *
+ * The nfsd_file object returned by this API is reference-counted
+ * and garbage-collected. The object is retained for a few
+ * seconds after the final nfsd_file_put() in case the caller
+ * wants to re-use it.
+ *
+ * Return values:
+ *   %nfs_ok - @pnf points to an nfsd_file with its reference
+ *   count boosted.
+ *
+ * On error, an nfsstat value in network byte order is returned.
+ */
+__be32
+nfsd_file_acquire_local(struct net *net, struct svc_cred *cred,
+			int nfs_vers, struct auth_domain *client,
+			struct svc_fh *fhp,
+			unsigned int may_flags, struct nfsd_file **pnf)
+{
+	return nfsd_file_do_acquire(NULL, net, cred, nfs_vers, client,
+				    fhp, may_flags, NULL, pnf, true);
 }
 
 /**
@@ -1193,7 +1241,8 @@ nfsd_file_acquire_opened(struct svc_rqst *rqstp, struct svc_fh *fhp,
 			 unsigned int may_flags, struct file *file,
 			 struct nfsd_file **pnf)
 {
-	return nfsd_file_do_acquire(rqstp, fhp, may_flags, file, pnf, false);
+	return nfsd_file_do_acquire(rqstp, SVC_NET(rqstp), NULL, 0, NULL,
+				    fhp, may_flags, file, pnf, false);
 }
 
 /*
fs/nfsd/filecache.h +4 -0
index 3fbec24eea6c..6dab41f8541e 100644
@@ -66,5 +66,9 @@ __be32 nfsd_file_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp,
 __be32 nfsd_file_acquire_opened(struct svc_rqst *rqstp, struct svc_fh *fhp,
 		  unsigned int may_flags, struct file *file,
 		  struct nfsd_file **nfp);
+__be32 nfsd_file_acquire_local(struct net *net, struct svc_cred *cred,
+			       int nfs_vers, struct auth_domain *client,
+			       struct svc_fh *fhp,
+			       unsigned int may_flags, struct nfsd_file **pnf);
 int nfsd_file_cache_stats_show(struct seq_file *m, void *v);
 #endif /* _FS_NFSD_FILECACHE_H */
fs/nfsd/nfsfh.c +0 -0
index 19e173187ab9..3635c0390cab 100644

fs/nfsd/nfsfh.c static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct net *net, · line 302

Quoted 3 lines
fs/nfsd/nfsfh.c +0 -0
@@ -300,7 +300,7 @@ static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct net *net,
 	return error;
 }
 

__fh_verify() becomes a public API here, so it now needs the kdoc comment (removed in 4/N) restored. Consider not removing that comment in 4/N.

Quoted 25 lines
fs/nfsd/nfsfh.c +1 -1
-static __be32
+__be32
 __fh_verify(struct svc_rqst *rqstp,
 	    struct net *net, struct svc_cred *cred,
 	    int nfs_vers, struct auth_domain *client,
fs/nfsd/nfsfh.h +5 -0
index 8d46e203d139..1429bee0ac1c 100644
@@ -217,6 +217,11 @@ extern char * SVCFH_fmt(struct svc_fh *fhp);
  * Function prototypes
  */
 __be32	fh_verify(struct svc_rqst *, struct svc_fh *, umode_t, int);
+__be32	__fh_verify(struct svc_rqst *rqstp,
+		    struct net *net, struct svc_cred *cred,
+		    int nfs_vers, struct auth_domain *client,
+		    struct auth_domain *gssclient,
+		    struct svc_fh *fhp, umode_t type, int access);
 __be32	fh_compose(struct svc_fh *, struct svc_export *, struct dentry *, struct svc_fh *);
 __be32	fh_update(struct svc_fh *);
 void	fh_put(struct svc_fh *);
Signature · 1 line
2.44.0
Signature · 1 line
Chuck Lever
#16 lore raw

PATCH v13 6/19 SUNRPC: remove call_allocate() BUG_ONs

In reply to Mike Snitzer

Remove BUG_ON if p_arglen=0 to allow RPC with void arg. Remove BUG_ON if p_replen=0 to allow RPC with void return.

The former was needed for the first revision of the LOCALIO protocol which had an RPC that took a void arg:

    /* raw RFC 9562 UUID */
    typedef u8 uuid_t<UUID_SIZE>;

    program NFS_LOCALIO_PROGRAM {
        version LOCALIO_V1 {
            void
                NULL(void) = 0;

            uuid_t
                GETUUID(void) = 1;
        } = 1;
    } = 400122;

The latter is needed for the final revision of the LOCALIO protocol which has a UUID_IS_LOCAL RPC which returns a void:

    /* raw RFC 9562 UUID */
    typedef u8 uuid_t<UUID_SIZE>;

    program NFS_LOCALIO_PROGRAM {
        version LOCALIO_V1 {
            void
                NULL(void) = 0;

            void
                UUID_IS_LOCAL(uuid_t) = 1;
        } = 1;
    } = 400122;

There is really no value in triggering a BUG_ON in response to either of these previously unsupported conditions.

NeilBrown would like the entire 'if (proc->p_proc != 0)' branch removed (not just the one BUG_ON that must be removed for LOCALIO's immediate needs of returning void).

---
 net/sunrpc/clnt.c | 6 ------
 1 file changed, 6 deletions(-)
net/sunrpc/clnt.c +0 -6
index 09f29a95f2bc..00fe6df11ab7 100644
@@ -1893,12 +1893,6 @@ call_allocate(struct rpc_task *task)
 	if (req->rq_buffer)
 		return;
 
-	if (proc->p_proc != 0) {
-		BUG_ON(proc->p_arglen == 0);
-		if (proc->p_decode != NULL)
-			BUG_ON(proc->p_replen == 0);
-	}
-
 	/*
 	 * Calculate the size (in quads) of the RPC call
 	 * and reply headers, and convert both values
Signature · 1 line
2.44.0
#17 lore raw

PATCH v13 7/19 SUNRPC: add rpcauth_map_clnt_to_svc_cred_local

In reply to Mike Snitzer

From: Weston Andros Adamson <[email protected]>

Add new funtion rpcauth_map_clnt_to_svc_cred_local which maps a generic cred to a svc_cred suitable for use in nfsd.

This is needed by the localio code to map nfs client creds to nfs server credentials.

Following from net/sunrpc/auth_unix.c:unx_marshal() it is clear that ->fsuid and ->fsgid must be used (rather than ->uid and ->gid). In addition, these uid and gid must be translated with from_kuid_munged() so local client uses correct uid and gid when acting as local server.

---
 include/linux/sunrpc/auth.h |  4 ++++
 net/sunrpc/auth.c           | 22 ++++++++++++++++++++++
 2 files changed, 26 insertions(+)
include/linux/sunrpc/auth.h +4 -0
index 61e58327b1aa..4cfb68f511db 100644
@@ -11,6 +11,7 @@
 #define _LINUX_SUNRPC_AUTH_H
 
 #include <linux/sunrpc/sched.h>
+#include <linux/sunrpc/svcauth.h>
 #include <linux/sunrpc/msg_prot.h>
 #include <linux/sunrpc/xdr.h>
 
@@ -184,6 +185,9 @@ int			rpcauth_uptodatecred(struct rpc_task *);
 int			rpcauth_init_credcache(struct rpc_auth *);
 void			rpcauth_destroy_credcache(struct rpc_auth *);
 void			rpcauth_clear_credcache(struct rpc_cred_cache *);
+void			rpcauth_map_clnt_to_svc_cred_local(struct rpc_clnt *clnt,
+							   const struct cred *,
+							   struct svc_cred *);
 char *			rpcauth_stringify_acceptor(struct rpc_cred *);
 
 static inline
net/sunrpc/auth.c +22 -0
index 04534ea537c8..3b6d91b36589 100644
@@ -17,6 +17,7 @@
 #include <linux/sunrpc/clnt.h>
 #include <linux/sunrpc/gss_api.h>
 #include <linux/spinlock.h>
+#include <linux/user_namespace.h>
 
 #include <trace/events/sunrpc.h>
 
@@ -308,6 +309,27 @@ rpcauth_init_credcache(struct rpc_auth *auth)
 }
 EXPORT_SYMBOL_GPL(rpcauth_init_credcache);
 
+void
+rpcauth_map_clnt_to_svc_cred_local(struct rpc_clnt *clnt,
+				   const struct cred *cred,
+				   struct svc_cred *svc)
+{
+	struct user_namespace *userns = clnt->cl_cred ?
+		clnt->cl_cred->user_ns : &init_user_ns;
+
+	memset(svc, 0, sizeof(struct svc_cred));
+
+	svc->cr_uid = KUIDT_INIT(from_kuid_munged(userns, cred->fsuid));
+	svc->cr_gid = KGIDT_INIT(from_kgid_munged(userns, cred->fsgid));
+	svc->cr_flavor = clnt->cl_auth->au_flavor;
+	if (cred->group_info)
+		svc->cr_group_info = get_group_info(cred->group_info);
+	/* These aren't relevant for local (network is bypassed) */
+	svc->cr_principal = NULL;
+	svc->cr_gss_mech = NULL;
+}
+EXPORT_SYMBOL_GPL(rpcauth_map_clnt_to_svc_cred_local);
+
 char *
 rpcauth_stringify_acceptor(struct rpc_cred *cred)
 {
Signature · 1 line
2.44.0
#18 lore raw

Re: [PATCH v13 07/19] SUNRPC: add rpcauth_map_clnt_to_svc_cred_local

Quoted 58 lines · Mike Snitzer

From: Weston Andros Adamson <[email protected]>

Add new funtion rpcauth_map_clnt_to_svc_cred_local which maps a generic cred to a svc_cred suitable for use in nfsd.

This is needed by the localio code to map nfs client creds to nfs server credentials.

Following from net/sunrpc/auth_unix.c:unx_marshal() it is clear that ->fsuid and ->fsgid must be used (rather than ->uid and ->gid). In addition, these uid and gid must be translated with from_kuid_munged() so local client uses correct uid and gid when acting as local server.

---
 include/linux/sunrpc/auth.h |  4 ++++
 net/sunrpc/auth.c           | 22 ++++++++++++++++++++++
 2 files changed, 26 insertions(+)
include/linux/sunrpc/auth.h +4 -0
index 61e58327b1aa..4cfb68f511db 100644
@@ -11,6 +11,7 @@
 #define _LINUX_SUNRPC_AUTH_H
 
 #include <linux/sunrpc/sched.h>
+#include <linux/sunrpc/svcauth.h>
 #include <linux/sunrpc/msg_prot.h>
 #include <linux/sunrpc/xdr.h>
 
@@ -184,6 +185,9 @@ int			rpcauth_uptodatecred(struct rpc_task *);
 int			rpcauth_init_credcache(struct rpc_auth *);
 void			rpcauth_destroy_credcache(struct rpc_auth *);
 void			rpcauth_clear_credcache(struct rpc_cred_cache *);
+void			rpcauth_map_clnt_to_svc_cred_local(struct rpc_clnt *clnt,
+							   const struct cred *,
+							   struct svc_cred *);
 char *			rpcauth_stringify_acceptor(struct rpc_cred *);
 
 static inline
net/sunrpc/auth.c +1 -0
index 04534ea537c8..3b6d91b36589 100644
@@ -17,6 +17,7 @@
 #include <linux/sunrpc/clnt.h>
 #include <linux/sunrpc/gss_api.h>
 #include <linux/spinlock.h>
+#include <linux/user_namespace.h>
 
 #include <trace/events/sunrpc.h>
 

net/sunrpc/auth.c rpcauth_init_credcache(struct rpc_auth *auth) · line 311

Quoted 3 lines
net/sunrpc/auth.c +0 -0
@@ -308,6 +309,27 @@ rpcauth_init_credcache(struct rpc_auth *auth)
 }
 EXPORT_SYMBOL_GPL(rpcauth_init_credcache);
 

rpcauth_map_clnt_to_svc_cred_local() needs a kdoc comment.

Since it is called only from fs/nfsd/localio.c -- should this API

                                ^^^^

reside in net/sunrpc/svcauth.c instead of net/sunrpc/auth.c ?

Quoted 27 lines
net/sunrpc/auth.c +21 -0
+void
+rpcauth_map_clnt_to_svc_cred_local(struct rpc_clnt *clnt,
+				   const struct cred *cred,
+				   struct svc_cred *svc)
+{
+	struct user_namespace *userns = clnt->cl_cred ?
+		clnt->cl_cred->user_ns : &init_user_ns;
+
+	memset(svc, 0, sizeof(struct svc_cred));
+
+	svc->cr_uid = KUIDT_INIT(from_kuid_munged(userns, cred->fsuid));
+	svc->cr_gid = KGIDT_INIT(from_kgid_munged(userns, cred->fsgid));
+	svc->cr_flavor = clnt->cl_auth->au_flavor;
+	if (cred->group_info)
+		svc->cr_group_info = get_group_info(cred->group_info);
+	/* These aren't relevant for local (network is bypassed) */
+	svc->cr_principal = NULL;
+	svc->cr_gss_mech = NULL;
+}
+EXPORT_SYMBOL_GPL(rpcauth_map_clnt_to_svc_cred_local);
+
 char *
 rpcauth_stringify_acceptor(struct rpc_cred *cred)
 {
Signature · 1 line
2.44.0
Signature · 1 line
Chuck Lever
#19 lore raw

Re: [PATCH v13 07/19] SUNRPC: add rpcauth_map_clnt_to_svc_cred_local

Quoted 65 lines · Chuck Lever
Quoted 58 lines · Mike Snitzer

From: Weston Andros Adamson <[email protected]>

Add new funtion rpcauth_map_clnt_to_svc_cred_local which maps a generic cred to a svc_cred suitable for use in nfsd.

This is needed by the localio code to map nfs client creds to nfs server credentials.

Following from net/sunrpc/auth_unix.c:unx_marshal() it is clear that ->fsuid and ->fsgid must be used (rather than ->uid and ->gid). In addition, these uid and gid must be translated with from_kuid_munged() so local client uses correct uid and gid when acting as local server.

---
 include/linux/sunrpc/auth.h |  4 ++++
 net/sunrpc/auth.c           | 22 ++++++++++++++++++++++
 2 files changed, 26 insertions(+)
include/linux/sunrpc/auth.h +4 -0
index 61e58327b1aa..4cfb68f511db 100644
@@ -11,6 +11,7 @@
 #define _LINUX_SUNRPC_AUTH_H
 
 #include <linux/sunrpc/sched.h>
+#include <linux/sunrpc/svcauth.h>
 #include <linux/sunrpc/msg_prot.h>
 #include <linux/sunrpc/xdr.h>
 
@@ -184,6 +185,9 @@ int			rpcauth_uptodatecred(struct rpc_task *);
 int			rpcauth_init_credcache(struct rpc_auth *);
 void			rpcauth_destroy_credcache(struct rpc_auth *);
 void			rpcauth_clear_credcache(struct rpc_cred_cache *);
+void			rpcauth_map_clnt_to_svc_cred_local(struct rpc_clnt *clnt,
+							   const struct cred *,
+							   struct svc_cred *);
 char *			rpcauth_stringify_acceptor(struct rpc_cred *);
 
 static inline
net/sunrpc/auth.c +1 -0
index 04534ea537c8..3b6d91b36589 100644
@@ -17,6 +17,7 @@
 #include <linux/sunrpc/clnt.h>
 #include <linux/sunrpc/gss_api.h>
 #include <linux/spinlock.h>
+#include <linux/user_namespace.h>
 
 #include <trace/events/sunrpc.h>
 

net/sunrpc/auth.c rpcauth_init_credcache(struct rpc_auth *auth) · line 311

Quoted 3 lines
net/sunrpc/auth.c +0 -0
@@ -308,6 +309,27 @@ rpcauth_init_credcache(struct rpc_auth *auth)
 }
 EXPORT_SYMBOL_GPL(rpcauth_init_credcache);
 

rpcauth_map_clnt_to_svc_cred_local() needs a kdoc comment.

Quoted 4 lines

Since it is called only from fs/nfsd/localio.c -- should this API

                                ^^^^

reside in net/sunrpc/svcauth.c instead of net/sunrpc/auth.c ?

Yes, that makes sense. I also renamed it to
svcauth_map_clnt_to_svc_cred_local

Thanks.

#20 lore raw

PATCH v13 8/19 SUNRPC: replace program list with program array

In reply to Mike Snitzer

From: NeilBrown <[email protected]>

A service created with svc_create_pooled() can be given a linked list of programs and all of these will be served.

Using a linked list makes it cumbersome when there are several programs that can be optionally selected with CONFIG settings.

After this patch is applied, API consumers must use only svc_create_pooled() when creating an RPC service that listens for more than one RPC program.

---
 fs/nfsd/nfsctl.c           |  2 +-
 fs/nfsd/nfsd.h             |  2 +-
 fs/nfsd/nfssvc.c           | 38 ++++++++++-----------
 include/linux/sunrpc/svc.h |  7 ++--
 net/sunrpc/svc.c           | 68 ++++++++++++++++++++++----------------
 net/sunrpc/svc_xprt.c      |  2 +-
 net/sunrpc/svcauth_unix.c  |  3 +-
 7 files changed, 67 insertions(+), 55 deletions(-)
fs/nfsd/nfsctl.c +1 -1
index 1c9e5b4bcb0a..64c1b4d649bc 100644
@@ -2246,7 +2246,7 @@ static __net_init int nfsd_net_init(struct net *net)
 	if (retval)
 		goto out_repcache_error;
 	memset(&nn->nfsd_svcstats, 0, sizeof(nn->nfsd_svcstats));
-	nn->nfsd_svcstats.program = &nfsd_program;
+	nn->nfsd_svcstats.program = &nfsd_programs[0];
 	for (i = 0; i < sizeof(nn->nfsd_versions); i++)
 		nn->nfsd_versions[i] = nfsd_support_version(i);
 	for (i = 0; i < sizeof(nn->nfsd4_minorversions); i++)
fs/nfsd/nfsd.h +1 -1
index 4ccbf014a2c7..b0d3e82d6dcd 100644
@@ -85,7 +85,7 @@ struct nfsd_genl_rqstp {
 	u32			rq_opnum[NFSD_MAX_OPS_PER_COMPOUND];
 };
 
-extern struct svc_program	nfsd_program;
+extern struct svc_program	nfsd_programs[];
 extern const struct svc_version	nfsd_version2, nfsd_version3, nfsd_version4;
 extern struct mutex		nfsd_mutex;
 extern spinlock_t		nfsd_drc_lock;
fs/nfsd/nfssvc.c +19 -19
index defc430f912f..b02eaaea7d62 100644
@@ -35,7 +35,6 @@
 #define NFSDDBG_FACILITY	NFSDDBG_SVC
 
 atomic_t			nfsd_th_cnt = ATOMIC_INIT(0);
-extern struct svc_program	nfsd_program;
 static int			nfsd(void *vrqstp);
 #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
 static int			nfsd_acl_rpcbind_set(struct net *,
@@ -90,20 +89,9 @@ static const struct svc_version *nfsd_acl_version[] = {
 # endif
 };
 
-#define NFSD_ACL_MINVERS            2
+#define NFSD_ACL_MINVERS	2
 #define NFSD_ACL_NRVERS		ARRAY_SIZE(nfsd_acl_version)
 
-static struct svc_program	nfsd_acl_program = {
-	.pg_prog		= NFS_ACL_PROGRAM,
-	.pg_nvers		= NFSD_ACL_NRVERS,
-	.pg_vers		= nfsd_acl_version,
-	.pg_name		= "nfsacl",
-	.pg_class		= "nfsd",
-	.pg_authenticate	= &svc_set_client,
-	.pg_init_request	= nfsd_acl_init_request,
-	.pg_rpcbind_set		= nfsd_acl_rpcbind_set,
-};
-
 #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */
 
 static const struct svc_version *nfsd_version[NFSD_MAXVERS+1] = {
@@ -116,18 +104,29 @@ static const struct svc_version *nfsd_version[NFSD_MAXVERS+1] = {
 #endif
 };
 
-struct svc_program		nfsd_program = {
-#if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
-	.pg_next		= &nfsd_acl_program,
-#endif
+struct svc_program		nfsd_programs[] = {
+	{
 	.pg_prog		= NFS_PROGRAM,		/* program number */
 	.pg_nvers		= NFSD_MAXVERS+1,	/* nr of entries in nfsd_version */
 	.pg_vers		= nfsd_version,		/* version table */
 	.pg_name		= "nfsd",		/* program name */
 	.pg_class		= "nfsd",		/* authentication class */
-	.pg_authenticate	= &svc_set_client,	/* export authentication */
+	.pg_authenticate	= svc_set_client,	/* export authentication */
 	.pg_init_request	= nfsd_init_request,
 	.pg_rpcbind_set		= nfsd_rpcbind_set,
+	},
+#if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
+	{
+	.pg_prog		= NFS_ACL_PROGRAM,
+	.pg_nvers		= NFSD_ACL_NRVERS,
+	.pg_vers		= nfsd_acl_version,
+	.pg_name		= "nfsacl",
+	.pg_class		= "nfsd",
+	.pg_authenticate	= svc_set_client,
+	.pg_init_request	= nfsd_acl_init_request,
+	.pg_rpcbind_set		= nfsd_acl_rpcbind_set,
+	},
+#endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */
 };
 
 bool nfsd_support_version(int vers)
@@ -598,7 +597,8 @@ int nfsd_create_serv(struct net *net)
 	if (nfsd_max_blksize == 0)
 		nfsd_max_blksize = nfsd_get_default_max_blksize();
 	nfsd_reset_versions(nn);
-	serv = svc_create_pooled(&nfsd_program, &nn->nfsd_svcstats,
+	serv = svc_create_pooled(nfsd_programs, ARRAY_SIZE(nfsd_programs),
+				 &nn->nfsd_svcstats,
 				 nfsd_max_blksize, nfsd);
 	if (serv == NULL)
 		return -ENOMEM;
include/linux/sunrpc/svc.h +4 -3
index 437672bcaa22..c7ad2fb2a155 100644
@@ -67,9 +67,10 @@ enum {
  * We currently do not support more than one RPC program per daemon.
  */
 struct svc_serv {
-	struct svc_program *	sv_program;	/* RPC program */
+	struct svc_program *	sv_programs;	/* RPC programs */
 	struct svc_stat *	sv_stats;	/* RPC statistics */
 	spinlock_t		sv_lock;
+	unsigned int		sv_nprogs;	/* Number of sv_programs */
 	unsigned int		sv_nrthreads;	/* # of server threads */
 	unsigned int		sv_maxconn;	/* max connections allowed or
 						 * '0' causing max to be based
@@ -357,10 +358,9 @@ struct svc_process_info {
 };
 
 /*
- * List of RPC programs on the same transport endpoint
+ * RPC program - an array of these can use the same transport endpoint
  */
 struct svc_program {
-	struct svc_program *	pg_next;	/* other programs (same xprt) */
 	u32			pg_prog;	/* program number */
 	unsigned int		pg_lovers;	/* lowest version */
 	unsigned int		pg_hivers;	/* highest version */
@@ -438,6 +438,7 @@ bool		   svc_rqst_replace_page(struct svc_rqst *rqstp,
 void		   svc_rqst_release_pages(struct svc_rqst *rqstp);
 void		   svc_exit_thread(struct svc_rqst *);
 struct svc_serv *  svc_create_pooled(struct svc_program *prog,
+				     unsigned int nprog,
 				     struct svc_stat *stats,
 				     unsigned int bufsize,
 				     int (*threadfn)(void *data));
net/sunrpc/svc.c +39 -29

166 lines not on this page · raw message

net/sunrpc/svc_xprt.c +1 -1
index 53ebc719ff5a..43c57124de52 100644
@@ -268,7 +268,7 @@ static int _svc_xprt_create(struct svc_serv *serv, const char *xprt_name,
 		spin_unlock(&svc_xprt_class_lock);
 		newxprt = xcl->xcl_ops->xpo_create(serv, net, sap, len, flags);
 		if (IS_ERR(newxprt)) {
-			trace_svc_xprt_create_err(serv->sv_program->pg_name,
+			trace_svc_xprt_create_err(serv->sv_programs->pg_name,
 						  xcl->xcl_name, sap, len,
 						  newxprt);
 			module_put(xcl->xcl_owner);
net/sunrpc/svcauth_unix.c +2 -1
index 04b45588ae6f..8ca98b146ec8 100644
@@ -697,7 +697,8 @@ svcauth_unix_set_client(struct svc_rqst *rqstp)
 	rqstp->rq_auth_stat = rpc_autherr_badcred;
 	ipm = ip_map_cached_get(xprt);
 	if (ipm == NULL)
-		ipm = __ip_map_lookup(sn->ip_map_cache, rqstp->rq_server->sv_program->pg_class,
+		ipm = __ip_map_lookup(sn->ip_map_cache,
+				      rqstp->rq_server->sv_programs->pg_class,
 				    &sin6->sin6_addr);
 
 	if (ipm == NULL)
Signature · 1 line
2.44.0
Acked-by on v13 8/19
#21 lore raw

Re: [PATCH v13 08/19] SUNRPC: replace program list with program array

Quoted 8 lines · Mike Snitzer

From: NeilBrown <[email protected]>

A service created with svc_create_pooled() can be given a linked list of programs and all of these will be served.

Using a linked list makes it cumbersome when there are several programs that can be optionally selected with CONFIG settings.

Quoted 6 lines

After this patch is applied, API consumers must use only svc_create_pooled() when creating an RPC service that listens for more than one RPC program.

There are a number of minor nits that I won't bother with here, but overall this is wholesome goodness.

Quoted 363 lines
---
 fs/nfsd/nfsctl.c           |  2 +-
 fs/nfsd/nfsd.h             |  2 +-
 fs/nfsd/nfssvc.c           | 38 ++++++++++-----------
 include/linux/sunrpc/svc.h |  7 ++--
 net/sunrpc/svc.c           | 68 ++++++++++++++++++++++----------------
 net/sunrpc/svc_xprt.c      |  2 +-
 net/sunrpc/svcauth_unix.c  |  3 +-
 7 files changed, 67 insertions(+), 55 deletions(-)
fs/nfsd/nfsctl.c +1 -1
index 1c9e5b4bcb0a..64c1b4d649bc 100644
@@ -2246,7 +2246,7 @@ static __net_init int nfsd_net_init(struct net *net)
 	if (retval)
 		goto out_repcache_error;
 	memset(&nn->nfsd_svcstats, 0, sizeof(nn->nfsd_svcstats));
-	nn->nfsd_svcstats.program = &nfsd_program;
+	nn->nfsd_svcstats.program = &nfsd_programs[0];
 	for (i = 0; i < sizeof(nn->nfsd_versions); i++)
 		nn->nfsd_versions[i] = nfsd_support_version(i);
 	for (i = 0; i < sizeof(nn->nfsd4_minorversions); i++)
fs/nfsd/nfsd.h +1 -1
index 4ccbf014a2c7..b0d3e82d6dcd 100644
@@ -85,7 +85,7 @@ struct nfsd_genl_rqstp {
 	u32			rq_opnum[NFSD_MAX_OPS_PER_COMPOUND];
 };
 
-extern struct svc_program	nfsd_program;
+extern struct svc_program	nfsd_programs[];
 extern const struct svc_version	nfsd_version2, nfsd_version3, nfsd_version4;
 extern struct mutex		nfsd_mutex;
 extern spinlock_t		nfsd_drc_lock;
fs/nfsd/nfssvc.c +19 -19
index defc430f912f..b02eaaea7d62 100644
@@ -35,7 +35,6 @@
 #define NFSDDBG_FACILITY	NFSDDBG_SVC
 
 atomic_t			nfsd_th_cnt = ATOMIC_INIT(0);
-extern struct svc_program	nfsd_program;
 static int			nfsd(void *vrqstp);
 #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
 static int			nfsd_acl_rpcbind_set(struct net *,
@@ -90,20 +89,9 @@ static const struct svc_version *nfsd_acl_version[] = {
 # endif
 };
 
-#define NFSD_ACL_MINVERS            2
+#define NFSD_ACL_MINVERS	2
 #define NFSD_ACL_NRVERS		ARRAY_SIZE(nfsd_acl_version)
 
-static struct svc_program	nfsd_acl_program = {
-	.pg_prog		= NFS_ACL_PROGRAM,
-	.pg_nvers		= NFSD_ACL_NRVERS,
-	.pg_vers		= nfsd_acl_version,
-	.pg_name		= "nfsacl",
-	.pg_class		= "nfsd",
-	.pg_authenticate	= &svc_set_client,
-	.pg_init_request	= nfsd_acl_init_request,
-	.pg_rpcbind_set		= nfsd_acl_rpcbind_set,
-};
-
 #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */
 
 static const struct svc_version *nfsd_version[NFSD_MAXVERS+1] = {
@@ -116,18 +104,29 @@ static const struct svc_version *nfsd_version[NFSD_MAXVERS+1] = {
 #endif
 };
 
-struct svc_program		nfsd_program = {
-#if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
-	.pg_next		= &nfsd_acl_program,
-#endif
+struct svc_program		nfsd_programs[] = {
+	{
 	.pg_prog		= NFS_PROGRAM,		/* program number */
 	.pg_nvers		= NFSD_MAXVERS+1,	/* nr of entries in nfsd_version */
 	.pg_vers		= nfsd_version,		/* version table */
 	.pg_name		= "nfsd",		/* program name */
 	.pg_class		= "nfsd",		/* authentication class */
-	.pg_authenticate	= &svc_set_client,	/* export authentication */
+	.pg_authenticate	= svc_set_client,	/* export authentication */
 	.pg_init_request	= nfsd_init_request,
 	.pg_rpcbind_set		= nfsd_rpcbind_set,
+	},
+#if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
+	{
+	.pg_prog		= NFS_ACL_PROGRAM,
+	.pg_nvers		= NFSD_ACL_NRVERS,
+	.pg_vers		= nfsd_acl_version,
+	.pg_name		= "nfsacl",
+	.pg_class		= "nfsd",
+	.pg_authenticate	= svc_set_client,
+	.pg_init_request	= nfsd_acl_init_request,
+	.pg_rpcbind_set		= nfsd_acl_rpcbind_set,
+	},
+#endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */
 };
 
 bool nfsd_support_version(int vers)
@@ -598,7 +597,8 @@ int nfsd_create_serv(struct net *net)
 	if (nfsd_max_blksize == 0)
 		nfsd_max_blksize = nfsd_get_default_max_blksize();
 	nfsd_reset_versions(nn);
-	serv = svc_create_pooled(&nfsd_program, &nn->nfsd_svcstats,
+	serv = svc_create_pooled(nfsd_programs, ARRAY_SIZE(nfsd_programs),
+				 &nn->nfsd_svcstats,
 				 nfsd_max_blksize, nfsd);
 	if (serv == NULL)
 		return -ENOMEM;
include/linux/sunrpc/svc.h +4 -3
index 437672bcaa22..c7ad2fb2a155 100644
@@ -67,9 +67,10 @@ enum {
  * We currently do not support more than one RPC program per daemon.
  */
 struct svc_serv {
-	struct svc_program *	sv_program;	/* RPC program */
+	struct svc_program *	sv_programs;	/* RPC programs */
 	struct svc_stat *	sv_stats;	/* RPC statistics */
 	spinlock_t		sv_lock;
+	unsigned int		sv_nprogs;	/* Number of sv_programs */
 	unsigned int		sv_nrthreads;	/* # of server threads */
 	unsigned int		sv_maxconn;	/* max connections allowed or
 						 * '0' causing max to be based
@@ -357,10 +358,9 @@ struct svc_process_info {
 };
 
 /*
- * List of RPC programs on the same transport endpoint
+ * RPC program - an array of these can use the same transport endpoint
  */
 struct svc_program {
-	struct svc_program *	pg_next;	/* other programs (same xprt) */
 	u32			pg_prog;	/* program number */
 	unsigned int		pg_lovers;	/* lowest version */
 	unsigned int		pg_hivers;	/* highest version */
@@ -438,6 +438,7 @@ bool		   svc_rqst_replace_page(struct svc_rqst *rqstp,
 void		   svc_rqst_release_pages(struct svc_rqst *rqstp);
 void		   svc_exit_thread(struct svc_rqst *);
 struct svc_serv *  svc_create_pooled(struct svc_program *prog,
+				     unsigned int nprog,
 				     struct svc_stat *stats,
 				     unsigned int bufsize,
 				     int (*threadfn)(void *data));
net/sunrpc/svc.c +39 -29
index ff6f3e35b36d..b33386d249c2 100644
@@ -440,10 +440,11 @@ EXPORT_SYMBOL_GPL(svc_rpcb_cleanup);
 
 static int svc_uses_rpcbind(struct svc_serv *serv)
 {
-	struct svc_program	*progp;
-	unsigned int		i;
+	unsigned int		p, i;
+
+	for (p = 0; p < serv->sv_nprogs; p++) {
+		struct svc_program *progp = &serv->sv_programs[p];
 
-	for (progp = serv->sv_program; progp; progp = progp->pg_next) {
 		for (i = 0; i < progp->pg_nvers; i++) {
 			if (progp->pg_vers[i] == NULL)
 				continue;
@@ -480,7 +481,7 @@ __svc_init_bc(struct svc_serv *serv)
  * Create an RPC service
  */
 static struct svc_serv *
-__svc_create(struct svc_program *prog, struct svc_stat *stats,
+__svc_create(struct svc_program *prog, int nprogs, struct svc_stat *stats,
 	     unsigned int bufsize, int npools, int (*threadfn)(void *data))
 {
 	struct svc_serv	*serv;
@@ -491,7 +492,8 @@ __svc_create(struct svc_program *prog, struct svc_stat *stats,
 	if (!(serv = kzalloc(sizeof(*serv), GFP_KERNEL)))
 		return NULL;
 	serv->sv_name      = prog->pg_name;
-	serv->sv_program   = prog;
+	serv->sv_programs  = prog;
+	serv->sv_nprogs    = nprogs;
 	serv->sv_stats     = stats;
 	if (bufsize > RPCSVC_MAXPAYLOAD)
 		bufsize = RPCSVC_MAXPAYLOAD;
@@ -499,17 +501,18 @@ __svc_create(struct svc_program *prog, struct svc_stat *stats,
 	serv->sv_max_mesg  = roundup(serv->sv_max_payload + PAGE_SIZE, PAGE_SIZE);
 	serv->sv_threadfn = threadfn;
 	xdrsize = 0;
-	while (prog) {
-		prog->pg_lovers = prog->pg_nvers-1;
-		for (vers=0; vers<prog->pg_nvers ; vers++)
-			if (prog->pg_vers[vers]) {
-				prog->pg_hivers = vers;
-				if (prog->pg_lovers > vers)
-					prog->pg_lovers = vers;
-				if (prog->pg_vers[vers]->vs_xdrsize > xdrsize)
-					xdrsize = prog->pg_vers[vers]->vs_xdrsize;
+	for (i = 0; i < nprogs; i++) {
+		struct svc_program *progp = &prog[i];
+
+		progp->pg_lovers = progp->pg_nvers-1;
+		for (vers = 0; vers < progp->pg_nvers ; vers++)
+			if (progp->pg_vers[vers]) {
+				progp->pg_hivers = vers;
+				if (progp->pg_lovers > vers)
+					progp->pg_lovers = vers;
+				if (progp->pg_vers[vers]->vs_xdrsize > xdrsize)
+					xdrsize = progp->pg_vers[vers]->vs_xdrsize;
 			}
-		prog = prog->pg_next;
 	}
 	serv->sv_xdrsize   = xdrsize;
 	INIT_LIST_HEAD(&serv->sv_tempsocks);
@@ -558,13 +561,14 @@ __svc_create(struct svc_program *prog, struct svc_stat *stats,
 struct svc_serv *svc_create(struct svc_program *prog, unsigned int bufsize,
 			    int (*threadfn)(void *data))
 {
-	return __svc_create(prog, NULL, bufsize, 1, threadfn);
+	return __svc_create(prog, 1, NULL, bufsize, 1, threadfn);
 }
 EXPORT_SYMBOL_GPL(svc_create);
 
 /**
  * svc_create_pooled - Create an RPC service with pooled threads
- * @prog: the RPC program the new service will handle
+ * @prog:  Array of RPC programs the new service will handle
+ * @nprogs: Number of programs in the array
  * @stats: the stats struct if desired
  * @bufsize: maximum message size for @prog
  * @threadfn: a function to service RPC requests for @prog
@@ -572,6 +576,7 @@ EXPORT_SYMBOL_GPL(svc_create);
  * Returns an instantiated struct svc_serv object or NULL.
  */
 struct svc_serv *svc_create_pooled(struct svc_program *prog,
+				   unsigned int nprogs,
 				   struct svc_stat *stats,
 				   unsigned int bufsize,
 				   int (*threadfn)(void *data))
@@ -579,7 +584,7 @@ struct svc_serv *svc_create_pooled(struct svc_program *prog,
 	struct svc_serv *serv;
 	unsigned int npools = svc_pool_map_get();
 
-	serv = __svc_create(prog, stats, bufsize, npools, threadfn);
+	serv = __svc_create(prog, nprogs, stats, bufsize, npools, threadfn);
 	if (!serv)
 		goto out_err;
 	serv->sv_is_pooled = true;
@@ -602,16 +607,16 @@ svc_destroy(struct svc_serv **servp)
 
 	*servp = NULL;
 
-	dprintk("svc: svc_destroy(%s)\n", serv->sv_program->pg_name);
+	dprintk("svc: svc_destroy(%s)\n", serv->sv_programs->pg_name);
 	timer_shutdown_sync(&serv->sv_temptimer);
 
 	/*
 	 * Remaining transports at this point are not expected.
 	 */
 	WARN_ONCE(!list_empty(&serv->sv_permsocks),
-		  "SVC: permsocks remain for %s\n", serv->sv_program->pg_name);
+		  "SVC: permsocks remain for %s\n", serv->sv_programs->pg_name);
 	WARN_ONCE(!list_empty(&serv->sv_tempsocks),
-		  "SVC: tempsocks remain for %s\n", serv->sv_program->pg_name);
+		  "SVC: tempsocks remain for %s\n", serv->sv_programs->pg_name);
 
 	cache_clean_deferred(serv);
 
@@ -1149,15 +1154,16 @@ int svc_register(const struct svc_serv *serv, struct net *net,
 		 const int family, const unsigned short proto,
 		 const unsigned short port)
 {
-	struct svc_program	*progp;
-	unsigned int		i;
+	unsigned int		p, i;
 	int			error = 0;
 
 	WARN_ON_ONCE(proto == 0 && port == 0);
 	if (proto == 0 && port == 0)
 		return -EINVAL;
 
-	for (progp = serv->sv_program; progp; progp = progp->pg_next) {
+	for (p = 0; p < serv->sv_nprogs; p++) {
+		struct svc_program *progp = &serv->sv_programs[p];
+
 		for (i = 0; i < progp->pg_nvers; i++) {
 
 			error = progp->pg_rpcbind_set(net, progp, i,
@@ -1209,13 +1215,14 @@ static void __svc_unregister(struct net *net, const u32 program, const u32 versi
 static void svc_unregister(const struct svc_serv *serv, struct net *net)
 {
 	struct sighand_struct *sighand;
-	struct svc_program *progp;
 	unsigned long flags;
-	unsigned int i;
+	unsigned int p, i;
 
 	clear_thread_flag(TIF_SIGPENDING);
 
-	for (progp = serv->sv_program; progp; progp = progp->pg_next) {
+	for (p = 0; p < serv->sv_nprogs; p++) {
+		struct svc_program *progp = &serv->sv_programs[p];
+
 		for (i = 0; i < progp->pg_nvers; i++) {
 			if (progp->pg_vers[i] == NULL)
 				continue;
@@ -1321,7 +1328,7 @@ svc_process_common(struct svc_rqst *rqstp)
 	struct svc_process_info process;
 	enum svc_auth_status	auth_res;
 	unsigned int		aoffset;
-	int			rc;
+	int			pr, rc;
 	__be32			*p;
 
 	/* Will be turned off only when NFSv4 Sessions are used */
@@ -1345,9 +1352,12 @@ svc_process_common(struct svc_rqst *rqstp)
 	rqstp->rq_vers = be32_to_cpup(p++);
 	rqstp->rq_proc = be32_to_cpup(p);
 
-	for (progp = serv->sv_program; progp; progp = progp->pg_next)
+	for (pr = 0; pr < serv->sv_nprogs; pr++) {
+		progp = &serv->sv_programs[pr];
+
 		if (rqstp->rq_prog == progp->pg_prog)
 			break;
+	}
 
 	/*
 	 * Decode auth data, and add verifier to reply buffer.
net/sunrpc/svc_xprt.c +1 -1
index 53ebc719ff5a..43c57124de52 100644
@@ -268,7 +268,7 @@ static int _svc_xprt_create(struct svc_serv *serv, const char *xprt_name,
 		spin_unlock(&svc_xprt_class_lock);
 		newxprt = xcl->xcl_ops->xpo_create(serv, net, sap, len, flags);
 		if (IS_ERR(newxprt)) {
-			trace_svc_xprt_create_err(serv->sv_program->pg_name,
+			trace_svc_xprt_create_err(serv->sv_programs->pg_name,
 						  xcl->xcl_name, sap, len,
 						  newxprt);
 			module_put(xcl->xcl_owner);
net/sunrpc/svcauth_unix.c +2 -1
index 04b45588ae6f..8ca98b146ec8 100644
@@ -697,7 +697,8 @@ svcauth_unix_set_client(struct svc_rqst *rqstp)
 	rqstp->rq_auth_stat = rpc_autherr_badcred;
 	ipm = ip_map_cached_get(xprt);
 	if (ipm == NULL)
-		ipm = __ip_map_lookup(sn->ip_map_cache, rqstp->rq_server->sv_program->pg_class,
+		ipm = __ip_map_lookup(sn->ip_map_cache,
+				      rqstp->rq_server->sv_programs->pg_class,
 				    &sin6->sin6_addr);
 
 	if (ipm == NULL)
Signature · 1 line
2.44.0
Signature · 1 line
Chuck Lever
#22 lore raw

PATCH v13 9/19 nfs_common: add NFS LOCALIO auxiliary protocol enablement

In reply to Mike Snitzer

fs/nfs_common/nfslocalio.c provides interfaces that enable an NFS client to generate a nonce (single-use UUID) and associated short-lived nfs_uuid_t struct, register it with nfs_common for subsequent lookup and verification by the NFS server and if matched the NFS server populates members in the nfs_uuid_t struct.

nfs_common's nfs_uuids list is the basis for localio enablement, as such it has members that point to nfsd memory for direct use by the client (e.g. 'net' is the server's network namespace, through it the client can access nn->nfsd_serv with proper rcu read access).

Also, expose localio's required nfsd symbols to NFS client:
- Cache nfsd_open_local_fh symbol (defined in next commit) and other required nfsd symbols in a globally accessible 'nfs_to' nfs_to_nfsd_t struct.

- Introduce nfsd_file_file() wrapper that provides access to nfsd_file's backing file. Keeps nfsd_file structure opaque to NFS client (as suggested by Jeff Layton).

- The addition of nfsd_file_get, nfsd_file_put and nfsd_file_file symbols prepares for the NFS client to use nfsd_file for localio.

- Despite the use of indirect function calls, caching these nfsd symbols for use by the client offers a ~10% performance win (compared to always doing get+call+put) for high IOPS workloads.

---
 fs/nfs_common/Makefile     |   3 +
 fs/nfs_common/nfslocalio.c | 194 +++++++++++++++++++++++++++++++++++++
 fs/nfsd/filecache.c        |  25 +++++
 fs/nfsd/filecache.h        |   1 +
 include/linux/nfslocalio.h |  56 +++++++++++
 5 files changed, 279 insertions(+)
 create mode 100644 fs/nfs_common/nfslocalio.c
 create mode 100644 include/linux/nfslocalio.h
fs/nfs_common/Makefile +3 -0
index e58b01bb8dda..a5e54809701e 100644
@@ -6,6 +6,9 @@
 obj-$(CONFIG_NFS_ACL_SUPPORT) += nfs_acl.o
 nfs_acl-objs := nfsacl.o
 
+obj-$(CONFIG_NFS_COMMON_LOCALIO_SUPPORT) += nfs_localio.o
+nfs_localio-objs := nfslocalio.o
+
 obj-$(CONFIG_GRACE_PERIOD) += grace.o
 obj-$(CONFIG_NFS_V4_2_SSC_HELPER) += nfs_ssc.o
 
fs/nfs_common/nfslocalio.c new file +194 -0

194 lines not on this page · raw message

fs/nfsd/filecache.c +25 -0
index 94ecb9ed0ed1..2a79c45ca27a 100644
@@ -39,6 +39,7 @@
 #include <linux/fsnotify.h>
 #include <linux/seq_file.h>
 #include <linux/rhashtable.h>
+#include <linux/nfslocalio.h>
 
 #include "vfs.h"
 #include "nfsd.h"
@@ -345,6 +346,10 @@ nfsd_file_get(struct nfsd_file *nf)
 		return nf;
 	return NULL;
 }
+EXPORT_SYMBOL_GPL(nfsd_file_get);
+
+/* Compile time type checking, not used by anything */
+static nfs_to_nfsd_file_get_t __maybe_unused nfsd_file_get_typecheck = nfsd_file_get;
 
 /**
  * nfsd_file_put - put the reference to a nfsd_file
@@ -389,6 +394,26 @@ nfsd_file_put(struct nfsd_file *nf)
 	if (refcount_dec_and_test(&nf->nf_ref))
 		nfsd_file_free(nf);
 }
+EXPORT_SYMBOL_GPL(nfsd_file_put);
+
+/* Compile time type checking, not used by anything */
+static nfs_to_nfsd_file_put_t __maybe_unused nfsd_file_put_typecheck = nfsd_file_put;
+
+/**
+ * nfsd_file_file - get the backing file of an nfsd_file
+ * @nf: nfsd_file of which to access the backing file.
+ *
+ * Return backing file for @nf.
+ */
+struct file *
+nfsd_file_file(struct nfsd_file *nf)
+{
+	return nf->nf_file;
+}
+EXPORT_SYMBOL_GPL(nfsd_file_file);
+
+/* Compile time type checking, not used by anything */
+static nfs_to_nfsd_file_file_t __maybe_unused nfsd_file_file_typecheck = nfsd_file_file;
 
 static void
 nfsd_file_dispose_list(struct list_head *dispose)
fs/nfsd/filecache.h +1 -0
index 6dab41f8541e..ab8a4423edd9 100644
@@ -56,6 +56,7 @@ int nfsd_file_cache_start_net(struct net *net);
 void nfsd_file_cache_shutdown_net(struct net *net);
 void nfsd_file_put(struct nfsd_file *nf);
 struct nfsd_file *nfsd_file_get(struct nfsd_file *nf);
+struct file *nfsd_file_file(struct nfsd_file *nf);
 void nfsd_file_close_inode_sync(struct inode *inode);
 void nfsd_file_net_dispose(struct nfsd_net *nn);
 bool nfsd_file_is_cached(struct inode *inode);
include/linux/nfslocalio.h new file +56 -0

56 lines not on this page · raw message

Signature · 1 line
2.44.0
#23 lore raw

Re: [PATCH v13 09/19] nfs_common: add NFS LOCALIO auxiliary protocol enablement

Quoted 5 lines · Mike Snitzer

Also, expose localio's required nfsd symbols to NFS client:
- Cache nfsd_open_local_fh symbol (defined in next commit) and other required nfsd symbols in a globally accessible 'nfs_to' nfs_to_nfsd_t struct.

I'm not thrilled with the mechanism for getting these symbols.

I'd rather nfsd passed the symbols to nfs_uuid_is_local(), and it stored them somewhere that nfs can see them. No need for reference counting etc. If nfs/localio holds an auth_domain, then it implicitly holds a reference to the nfsd module and the functions cannot disappear.

I would created an 'nfs_localio_operations' structure which is defined in nfsd as a constant.
The address of this is passed to nfs_uud_is_local() and that address is stored in nfs_to if it doesn't already have the correct value.

So no need for symbol_request() or symbol_put().

Review

Quoted 2 lines
(quoted diff) +2 -0
+
+DEFINE_MUTEX(nfs_uuid_mutex);

This doesn't need to be a mutex - a spinlock is sufficient.

fs/nfs_common/nfslocalio.c line 63

Quoted 11 lines
fs/nfs_common/nfslocalio.c new file +11 -0
+
+bool nfs_uuid_is_local(const uuid_t *uuid, struct net *net, struct auth_domain *dom)
+{
+	bool is_local = false;
+	nfs_uuid_t *nfs_uuid;
+
+	rcu_read_lock();
+	nfs_uuid = nfs_uuid_lookup(uuid);
+	if (nfs_uuid) {
+		is_local = true;
+		nfs_uuid->net = net;

It looks odd that you don't take a reference to the net. It is probably correct but a comment explaining why would help. Is it that the dom implies a reference to the net?

fs/nfs_common/nfslocalio.c line 71

Quoted 8 lines
fs/nfs_common/nfslocalio.c new file +8 -0
+		kref_get(&dom->ref);
+		nfs_uuid->dom = dom;
+	}
+	rcu_read_unlock();
+
+	return is_local;
+}
+EXPORT_SYMBOL_GPL(nfs_uuid_is_local);

Thanks,
NeilBrown

#24 lore raw

Re: [PATCH v13 09/19] nfs_common: add NFS LOCALIO auxiliary protocol enablement

Quoted 14 lines · NeilBrown
Quoted 5 lines · Mike Snitzer

Also, expose localio's required nfsd symbols to NFS client:
- Cache nfsd_open_local_fh symbol (defined in next commit) and other required nfsd symbols in a globally accessible 'nfs_to' nfs_to_nfsd_t struct.

I'm not thrilled with the mechanism for getting these symbols.

I'd rather nfsd passed the symbols to nfs_uuid_is_local(), and it stored them somewhere that nfs can see them. No need for reference counting etc. If nfs/localio holds an auth_domain, then it implicitly holds a reference to the nfsd module and the functions cannot disappear.

Quoted 6 lines

I would created an 'nfs_localio_operations' structure which is defined in nfsd as a constant.
The address of this is passed to nfs_uud_is_local() and that address is stored in nfs_to if it doesn't already have the correct value.

So no need for symbol_request() or symbol_put().

I'm not seeing why we'd want to actively engineer some even more bespoke way to access nfsd symbols. The symbol refcounting is only done when the client first connects as part of the LOCALIO handshake (or when client is destroyed and LOCALIO enabled), so it isn't getting in the way.

Happy to revisit this but I'd really prefer to use standard convention (symbol_request and symbol_put) to establish nfs's dependency on nfsd's symbols.

Quoted 4 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+
+DEFINE_MUTEX(nfs_uuid_mutex);

This doesn't need to be a mutex - a spinlock is sufficient.

Fixed, thanks.

Quoted 7 lines
Quoted 7 lines
fs/nfs_common/nfslocalio.c new file +7 -0
+
+bool nfs_uuid_is_local(const uuid_t *uuid, struct net *net, struct auth_domain *dom)
+{
+	bool is_local = false;
+	nfs_uuid_t *nfs_uuid;
+
+	rcu_read_lock();
Quoted 8 lines

fs/nfs_common/nfslocalio.c line 63

Quoted 4 lines
fs/nfs_common/nfslocalio.c new file +4 -0
+	nfs_uuid = nfs_uuid_lookup(uuid);
+	if (nfs_uuid) {
+		is_local = true;
+		nfs_uuid->net = net;

It looks odd that you don't take a reference to the net. It is probably correct but a comment explaining why would help. Is it that the dom implies a reference to the net?

No, I just made the code tolerate the net having been torn down (see: fs/nfsd/localio.c:nfsd_open_local_fh), but it'd be safer to take a proper reference here.

I'll work through it once more and update accordingly.

Thanks!

#25 lore raw

Re: [PATCH v13 09/19] nfs_common: add NFS LOCALIO auxiliary protocol enablement

Quoted 28 lines · Mike Snitzer
Quoted 14 lines · NeilBrown
Quoted 5 lines · Mike Snitzer

Also, expose localio's required nfsd symbols to NFS client:
- Cache nfsd_open_local_fh symbol (defined in next commit) and other required nfsd symbols in a globally accessible 'nfs_to' nfs_to_nfsd_t struct.

I'm not thrilled with the mechanism for getting these symbols.

I'd rather nfsd passed the symbols to nfs_uuid_is_local(), and it stored them somewhere that nfs can see them. No need for reference counting etc. If nfs/localio holds an auth_domain, then it implicitly holds a reference to the nfsd module and the functions cannot disappear.

Quoted 6 lines

I would created an 'nfs_localio_operations' structure which is defined in nfsd as a constant.
The address of this is passed to nfs_uud_is_local() and that address is stored in nfs_to if it doesn't already have the correct value.

So no need for symbol_request() or symbol_put().

I'm not seeing why we'd want to actively engineer some even more bespoke way to access nfsd symbols. The symbol refcounting is only done when the client first connects as part of the LOCALIO handshake (or when client is destroyed and LOCALIO enabled), so it isn't getting in the way.

Quoted 3 lines

Happy to revisit this but I'd really prefer to use standard convention (symbol_request and symbol_put) to establish nfs's dependency on nfsd's symbols.

symbol_request is used about 21 times in the whole kernel. Given how much interaction there is between modules, that seems like a niche, not a standard :-)

From my perspective the "standard" way for code to call into other modules is for the module to provide an "operations" structure which it registers with the caller somehow. There are 48 instances of
struct .*_operations {
in include/. There are 494 with just _ops. 21 < 48+494

The sense in which refcounting gets in the way is not at runtime as I think you defend. It is at review time. Declaring an operations structure is trivial to review and doesn't require any clever macros (clever macros certainly have their place, but I don't think it is here).

There would be less code with an operations structure, so less review.

Thanks,
NeilBrown

#26 lore raw

PATCH v13 10/19 nfsd: add localio support

In reply to Mike Snitzer

From: Weston Andros Adamson <[email protected]>

Add server support for bypassing NFS for localhost reads, writes, and commits. This is only useful when both the client and server are running on the same host.

If nfsd_open_local_fh() fails then the NFS client will both retry and fallback to normal network-based read, write and commit operations if localio is no longer supported.

Care is taken to ensure the same NFS security mechanisms are used (authentication, etc) regardless of whether localio or regular NFS access is used. The auth_domain established as part of the traditional NFS client access to the NFS server is also used for localio. Store auth_domain for localio in nfsd_uuid_t and transfer it to the client if it is local to the server.

Relative to containers, localio gives the client access to the network namespace the server has. This is required to allow the client to access the server's per-namespace nfsd_net struct.

CONFIG_NFSD_LOCALIO controls the server enablement for localio. A later commit will add CONFIG_NFS_LOCALIO to allow the client enablement.

This commit also introduces the use of a percpu_ref to interlock nfsd_destroy_serv and nfsd_open_local_fh, and warrants a more detailed explanation:

Introduce nfsd_serv_try_get and nfsd_serv_put and update the nfsd code to prevent nfsd_destroy_serv from destroying nn->nfsd_serv until any client initiated localio calls to nfsd (that are _not_ in the context of nfsd) are complete.

nfsd_open_local_fh is updated to nfsd_serv_try_get before opening its file handle and then drop the reference using nfsd_serv_put at the end of nfsd_open_local_fh.

This "interlock" working relies heavily on nfsd_open_local_fh()'s maybe_get_net() safely dealing with the possibility that the struct net (and nfsd_net by association) may have been destroyed by nfsd_destroy_serv() via nfsd_shutdown_net().

Verified to fix an easy to hit crash that would occur if an nfsd instance running in a container, with a localio client mounted, is shutdown. Upon restart of the container and associated nfsd the client would go on to crash due to NULL pointer dereference that occuured due to the nfs client's localio attempting to nfsd_open_local_fh(), using nn->nfsd_serv, without having a proper reference on nn->nfsd_serv.

---
 fs/Kconfig          |   3 ++
 fs/nfsd/Kconfig     |  14 ++++++
 fs/nfsd/Makefile    |   1 +
 fs/nfsd/filecache.c |   2 +-
 fs/nfsd/localio.c   | 108 ++++++++++++++++++++++++++++++++++++++++++++
 fs/nfsd/netns.h     |   8 +++-
 fs/nfsd/nfssvc.c    |  39 ++++++++++++++++
 fs/nfsd/trace.h     |   3 +-
 fs/nfsd/vfs.h       |  10 ++++
 9 files changed, 185 insertions(+), 3 deletions(-)
 create mode 100644 fs/nfsd/localio.c
fs/Kconfig +3 -0
index a46b0cbc4d8f..1b8a5edbddff 100644
@@ -377,6 +377,9 @@ config NFS_ACL_SUPPORT
 	tristate
 	select FS_POSIX_ACL
 
+config NFS_COMMON_LOCALIO_SUPPORT
+	bool
+
 config NFS_COMMON
 	bool
 	depends on NFSD || NFS_FS || LOCKD
fs/nfsd/Kconfig +14 -0
index c0bd1509ccd4..1fca57c79c60 100644
@@ -90,6 +90,20 @@ config NFSD_V4
 
 	  If unsure, say N.
 
+config NFSD_LOCALIO
+	bool "NFS server support for the LOCALIO auxiliary protocol"
+	depends on NFSD
+	select NFS_COMMON_LOCALIO_SUPPORT
+	help
+	  Some NFS servers support an auxiliary NFS LOCALIO protocol
+	  that is not an official part of the NFS protocol.
+
+	  This option enables support for the LOCALIO protocol in the
+	  kernel's NFS server.  Enable this to bypass using the NFS
+	  protocol when issuing reads, writes and commits to the server.
+
+	  If unsure, say N.
+
 config NFSD_PNFS
 	bool
 
fs/nfsd/Makefile +1 -0
index b8736a82e57c..78b421778a79 100644
@@ -23,3 +23,4 @@ nfsd-$(CONFIG_NFSD_PNFS) += nfs4layouts.o
 nfsd-$(CONFIG_NFSD_BLOCKLAYOUT) += blocklayout.o blocklayoutxdr.o
 nfsd-$(CONFIG_NFSD_SCSILAYOUT) += blocklayout.o blocklayoutxdr.o
 nfsd-$(CONFIG_NFSD_FLEXFILELAYOUT) += flexfilelayout.o flexfilelayoutxdr.o
+nfsd-$(CONFIG_NFSD_LOCALIO) += localio.o
fs/nfsd/filecache.c +1 -1
index 2a79c45ca27a..d7c6122231f4 100644
@@ -53,7 +53,7 @@
 #define NFSD_FILE_CACHE_UP		     (0)
 
 /* We only care about NFSD_MAY_READ/WRITE for this cache */
-#define NFSD_FILE_MAY_MASK	(NFSD_MAY_READ|NFSD_MAY_WRITE)
+#define NFSD_FILE_MAY_MASK	(NFSD_MAY_READ|NFSD_MAY_WRITE|NFSD_MAY_LOCALIO)
 
 static DEFINE_PER_CPU(unsigned long, nfsd_file_cache_hits);
 static DEFINE_PER_CPU(unsigned long, nfsd_file_acquisitions);
fs/nfsd/localio.c new file +108 -0

108 lines not on this page · raw message

fs/nfsd/netns.h +7 -1
index 238fc4e56e53..e2d953f21dde 100644
@@ -13,6 +13,7 @@
 #include <linux/filelock.h>
 #include <linux/nfs4.h>
 #include <linux/percpu_counter.h>
+#include <linux/percpu-refcount.h>
 #include <linux/siphash.h>
 #include <linux/sunrpc/stats.h>
 
@@ -139,7 +140,9 @@ struct nfsd_net {
 
 	struct svc_info nfsd_info;
 #define nfsd_serv nfsd_info.serv
-
+	struct percpu_ref nfsd_serv_ref;
+	struct completion nfsd_serv_confirm_done;
+	struct completion nfsd_serv_free_done;
 
 	/*
 	 * clientid and stateid data for construction of net unique COPY
@@ -221,6 +224,9 @@ struct nfsd_net {
 extern bool nfsd_support_version(int vers);
 extern unsigned int nfsd_net_id;
 
+bool nfsd_serv_try_get(struct nfsd_net *nn);
+void nfsd_serv_put(struct nfsd_net *nn);
+
 void nfsd_copy_write_verifier(__be32 verf[2], struct nfsd_net *nn);
 void nfsd_reset_write_verifier(struct nfsd_net *nn);
 #endif /* __NFSD_NETNS_H__ */
fs/nfsd/nfssvc.c +39 -0
index b02eaaea7d62..c639fbe4d8c2 100644
@@ -192,6 +192,30 @@ int nfsd_minorversion(struct nfsd_net *nn, u32 minorversion, enum vers_op change
 	return 0;
 }
 
+bool nfsd_serv_try_get(struct nfsd_net *nn)
+{
+	return percpu_ref_tryget_live(&nn->nfsd_serv_ref);
+}
+
+void nfsd_serv_put(struct nfsd_net *nn)
+{
+	percpu_ref_put(&nn->nfsd_serv_ref);
+}
+
+static void nfsd_serv_done(struct percpu_ref *ref)
+{
+	struct nfsd_net *nn = container_of(ref, struct nfsd_net, nfsd_serv_ref);
+
+	complete(&nn->nfsd_serv_confirm_done);
+}
+
+static void nfsd_serv_free(struct percpu_ref *ref)
+{
+	struct nfsd_net *nn = container_of(ref, struct nfsd_net, nfsd_serv_ref);
+
+	complete(&nn->nfsd_serv_free_done);
+}
+
 /*
  * Maximum number of nfsd processes
  */
@@ -391,6 +415,7 @@ static void nfsd_shutdown_net(struct net *net)
 		lockd_down(net);
 		nn->lockd_up = false;
 	}
+	percpu_ref_exit(&nn->nfsd_serv_ref);
 	nn->nfsd_net_up = false;
 	nfsd_shutdown_generic();
 }
@@ -470,6 +495,13 @@ void nfsd_destroy_serv(struct net *net)
 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 	struct svc_serv *serv = nn->nfsd_serv;
 
+	lockdep_assert_held(&nfsd_mutex);
+
+	percpu_ref_kill_and_confirm(&nn->nfsd_serv_ref, nfsd_serv_done);
+	wait_for_completion(&nn->nfsd_serv_confirm_done);
+	wait_for_completion(&nn->nfsd_serv_free_done);
+	/* percpu_ref_exit is called in nfsd_shutdown_net */
+
 	spin_lock(&nfsd_notifier_lock);
 	nn->nfsd_serv = NULL;
 	spin_unlock(&nfsd_notifier_lock);
@@ -594,6 +626,13 @@ int nfsd_create_serv(struct net *net)
 	if (nn->nfsd_serv)
 		return 0;
 
+	error = percpu_ref_init(&nn->nfsd_serv_ref, nfsd_serv_free,
+				0, GFP_KERNEL);
+	if (error)
+		return error;
+	init_completion(&nn->nfsd_serv_free_done);
+	init_completion(&nn->nfsd_serv_confirm_done);
+
 	if (nfsd_max_blksize == 0)
 		nfsd_max_blksize = nfsd_get_default_max_blksize();
 	nfsd_reset_versions(nn);
fs/nfsd/trace.h +2 -1
index 77bbd23aa150..9c0610fdd11c 100644
@@ -86,7 +86,8 @@ DEFINE_NFSD_XDR_ERR_EVENT(cant_encode);
 		{ NFSD_MAY_NOT_BREAK_LEASE,	"NOT_BREAK_LEASE" },	\
 		{ NFSD_MAY_BYPASS_GSS,		"BYPASS_GSS" },		\
 		{ NFSD_MAY_READ_IF_EXEC,	"READ_IF_EXEC" },	\
-		{ NFSD_MAY_64BIT_COOKIE,	"64BIT_COOKIE" })
+		{ NFSD_MAY_64BIT_COOKIE,	"64BIT_COOKIE" },	\
+		{ NFSD_MAY_LOCALIO,		"LOCALIO" })
 
 TRACE_EVENT(nfsd_compound,
 	TP_PROTO(
fs/nfsd/vfs.h +10 -0
index 01947561d375..ec8a8aae540b 100644
@@ -33,6 +33,8 @@
 
 #define NFSD_MAY_64BIT_COOKIE		0x1000 /* 64 bit readdir cookies for >= NFSv3 */
 
+#define NFSD_MAY_LOCALIO		0x2000 /* for tracing, reflects when localio used */
+
 #define NFSD_MAY_CREATE		(NFSD_MAY_EXEC|NFSD_MAY_WRITE)
 #define NFSD_MAY_REMOVE		(NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC)
 
@@ -158,6 +160,14 @@ __be32		nfsd_permission(struct svc_cred *cred, struct svc_export *exp,
 
 void		nfsd_filp_close(struct file *fp);
 
+int		nfsd_open_local_fh(struct net *net,
+				   struct auth_domain *dom,
+				   struct rpc_clnt *rpc_clnt,
+				   const struct cred *cred,
+				   const struct nfs_fh *nfs_fh,
+				   const fmode_t fmode,
+				   struct nfsd_file **pnf);
+
 static inline int fh_want_write(struct svc_fh *fh)
 {
 	int ret;
Signature · 1 line
2.44.0
#27 lore raw

Re: [PATCH v13 10/19] nfsd: add localio support

Quoted 86 lines · Mike Snitzer

From: Weston Andros Adamson <[email protected]>

Add server support for bypassing NFS for localhost reads, writes, and commits. This is only useful when both the client and server are running on the same host.

If nfsd_open_local_fh() fails then the NFS client will both retry and fallback to normal network-based read, write and commit operations if localio is no longer supported.

Care is taken to ensure the same NFS security mechanisms are used (authentication, etc) regardless of whether localio or regular NFS access is used. The auth_domain established as part of the traditional NFS client access to the NFS server is also used for localio. Store auth_domain for localio in nfsd_uuid_t and transfer it to the client if it is local to the server.

Relative to containers, localio gives the client access to the network namespace the server has. This is required to allow the client to access the server's per-namespace nfsd_net struct.

CONFIG_NFSD_LOCALIO controls the server enablement for localio. A later commit will add CONFIG_NFS_LOCALIO to allow the client enablement.

This commit also introduces the use of a percpu_ref to interlock nfsd_destroy_serv and nfsd_open_local_fh, and warrants a more detailed explanation:

Introduce nfsd_serv_try_get and nfsd_serv_put and update the nfsd code to prevent nfsd_destroy_serv from destroying nn->nfsd_serv until any client initiated localio calls to nfsd (that are _not_ in the context of nfsd) are complete.

nfsd_open_local_fh is updated to nfsd_serv_try_get before opening its file handle and then drop the reference using nfsd_serv_put at the end of nfsd_open_local_fh.

This "interlock" working relies heavily on nfsd_open_local_fh()'s maybe_get_net() safely dealing with the possibility that the struct net (and nfsd_net by association) may have been destroyed by nfsd_destroy_serv() via nfsd_shutdown_net().

Verified to fix an easy to hit crash that would occur if an nfsd instance running in a container, with a localio client mounted, is shutdown. Upon restart of the container and associated nfsd the client would go on to crash due to NULL pointer dereference that occuured due to the nfs client's localio attempting to nfsd_open_local_fh(), using nn->nfsd_serv, without having a proper reference on nn->nfsd_serv.

---
 fs/Kconfig          |   3 ++
 fs/nfsd/Kconfig     |  14 ++++++
 fs/nfsd/Makefile    |   1 +
 fs/nfsd/filecache.c |   2 +-
 fs/nfsd/localio.c   | 108 ++++++++++++++++++++++++++++++++++++++++++++
 fs/nfsd/netns.h     |   8 +++-
 fs/nfsd/nfssvc.c    |  39 ++++++++++++++++
 fs/nfsd/trace.h     |   3 +-
 fs/nfsd/vfs.h       |  10 ++++
 9 files changed, 185 insertions(+), 3 deletions(-)
 create mode 100644 fs/nfsd/localio.c
fs/Kconfig +3 -0
index a46b0cbc4d8f..1b8a5edbddff 100644
@@ -377,6 +377,9 @@ config NFS_ACL_SUPPORT
 	tristate
 	select FS_POSIX_ACL
 
+config NFS_COMMON_LOCALIO_SUPPORT
+	bool
+
 config NFS_COMMON
 	bool
 	depends on NFSD || NFS_FS || LOCKD
fs/nfsd/Kconfig +0 -0
index c0bd1509ccd4..1fca57c79c60 100644

fs/nfsd/Kconfig config NFSD_V4 · line 96

Quoted 7 lines
fs/nfsd/Kconfig +4 -0
@@ -90,6 +90,20 @@ config NFSD_V4
 
 	  If unsure, say N.
 
+config NFSD_LOCALIO
+	bool "NFS server support for the LOCALIO auxiliary protocol"
+	depends on NFSD
+	select NFS_COMMON_LOCALIO_SUPPORT

Because this is a new and experimental facility that is still under development, CONFIG_NFSD_LOCALIO should default N.

fs/nfsd/Kconfig config NFSD_V4 · line 103

Quoted 7 lines
fs/nfsd/Kconfig +7 -0
+	help
+	  Some NFS servers support an auxiliary NFS LOCALIO protocol
+	  that is not an official part of the NFS protocol.
+
+	  This option enables support for the LOCALIO protocol in the
+	  kernel's NFS server.  Enable this to bypass using the NFS
+	  protocol when issuing reads, writes and commits to the server.

I found the last sentence in this paragraph confusing. Consider
instead: “Enable this to permit local NFS clients to bypass the network when issuing reads and write to the local NFS server.”

Quoted 197 lines
fs/nfsd/Kconfig +3 -0
+
+	  If unsure, say N.
+
 config NFSD_PNFS
 	bool
 
fs/nfsd/Makefile +1 -0
index b8736a82e57c..78b421778a79 100644
@@ -23,3 +23,4 @@ nfsd-$(CONFIG_NFSD_PNFS) += nfs4layouts.o
 nfsd-$(CONFIG_NFSD_BLOCKLAYOUT) += blocklayout.o blocklayoutxdr.o
 nfsd-$(CONFIG_NFSD_SCSILAYOUT) += blocklayout.o blocklayoutxdr.o
 nfsd-$(CONFIG_NFSD_FLEXFILELAYOUT) += flexfilelayout.o flexfilelayoutxdr.o
+nfsd-$(CONFIG_NFSD_LOCALIO) += localio.o
fs/nfsd/filecache.c +1 -1
index 2a79c45ca27a..d7c6122231f4 100644
@@ -53,7 +53,7 @@
 #define NFSD_FILE_CACHE_UP		     (0)
 
 /* We only care about NFSD_MAY_READ/WRITE for this cache */
-#define NFSD_FILE_MAY_MASK	(NFSD_MAY_READ|NFSD_MAY_WRITE)
+#define NFSD_FILE_MAY_MASK	(NFSD_MAY_READ|NFSD_MAY_WRITE|NFSD_MAY_LOCALIO)
 
 static DEFINE_PER_CPU(unsigned long, nfsd_file_cache_hits);
 static DEFINE_PER_CPU(unsigned long, nfsd_file_acquisitions);
fs/nfsd/localio.c new file +108 -0
new file mode 100644
index 000000000000..8e0cce835f81
@@ -0,0 +1,108 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * NFS server support for local clients to bypass network stack
+ *
+ * Copyright (C) 2014 Weston Andros Adamson <[email protected]>
+ * Copyright (C) 2019 Trond Myklebust <[email protected]>
+ * Copyright (C) 2024 Mike Snitzer <[email protected]>
+ */
+
+#include <linux/exportfs.h>
+#include <linux/sunrpc/svcauth_gss.h>
+#include <linux/sunrpc/clnt.h>
+#include <linux/nfs.h>
+#include <linux/nfs_common.h>
+#include <linux/nfslocalio.h>
+#include <linux/string.h>
+
+#include "nfsd.h"
+#include "vfs.h"
+#include "netns.h"
+#include "filecache.h"
+
+/**
+ * nfsd_open_local_fh - lookup a local filehandle @nfs_fh and map to @file
+ *
+ * @cl_nfssvc_net: the 'struct net' to use to get the proper nfsd_net
+ * @cl_nfssvc_dom: the 'struct auth_domain' required for localio access
+ * @rpc_clnt: rpc_clnt that the client established, used for sockaddr and cred
+ * @cred: cred that the client established
+ * @nfs_fh: filehandle to lookup
+ * @fmode: fmode_t to use for open
+ * @pnf: returned nfsd_file pointer that maps to @nfs_fh
+ *
+ * This function maps a local fh to a path on a local filesystem.
+ * This is useful when the nfs client has the local server mounted - it can
+ * avoid all the NFS overhead with reads, writes and commits.
+ *
+ * On successful return, caller is responsible for calling nfsd_file_put. Also
+ * note that this is called from nfs.ko via find_symbol() to avoid an explicit
+ * dependency on knfsd. So, there is no forward declaration in a header file
+ * for it that is shared with the client.
+ */
+int nfsd_open_local_fh(struct net *cl_nfssvc_net,
+			 struct auth_domain *cl_nfssvc_dom,
+			 struct rpc_clnt *rpc_clnt,
+			 const struct cred *cred,
+			 const struct nfs_fh *nfs_fh,
+			 const fmode_t fmode,
+			 struct nfsd_file **pnf)
+{
+	int mayflags = NFSD_MAY_LOCALIO;
+	int status = 0;
+	struct nfsd_net *nn;
+	const struct cred *save_cred;
+	struct svc_cred rq_cred;
+	struct svc_fh fh;
+	__be32 beres;
+
+	if (nfs_fh->size > NFS4_FHSIZE)
+		return -EINVAL;
+
+	/* Not running in nfsd context, must safely get reference on nfsd_serv */
+	cl_nfssvc_net = maybe_get_net(cl_nfssvc_net);
+	if (!cl_nfssvc_net)
+		return -ENXIO;
+	nn = net_generic(cl_nfssvc_net, nfsd_net_id);
+
+	/* The server may already be shutting down, disallow new localio */
+	if (unlikely(!nfsd_serv_try_get(nn))) {
+		status = -ENXIO;
+		goto out_net;
+	}
+
+	/* Save client creds before calling nfsd_file_acquire_local which calls nfsd_setuser */
+	save_cred = get_current_cred();
+
+	/* nfs_fh -> svc_fh */
+	fh_init(&fh, NFS4_FHSIZE);
+	fh.fh_handle.fh_size = nfs_fh->size;
+	memcpy(fh.fh_handle.fh_raw, nfs_fh->data, nfs_fh->size);
+
+	if (fmode & FMODE_READ)
+		mayflags |= NFSD_MAY_READ;
+	if (fmode & FMODE_WRITE)
+		mayflags |= NFSD_MAY_WRITE;
+
+	rpcauth_map_clnt_to_svc_cred_local(rpc_clnt, cred, &rq_cred);
+
+	beres = nfsd_file_acquire_local(cl_nfssvc_net, &rq_cred, rpc_clnt->cl_vers,
+					cl_nfssvc_dom, &fh, mayflags, pnf);
+	if (beres) {
+		status = nfs_stat_to_errno(be32_to_cpu(beres));
+		goto out_fh_put;
+	}
+out_fh_put:
+	fh_put(&fh);
+	if (rq_cred.cr_group_info)
+		put_group_info(rq_cred.cr_group_info);
+	revert_creds(save_cred);
+	nfsd_serv_put(nn);
+out_net:
+	put_net(cl_nfssvc_net);
+	return status;
+}
+EXPORT_SYMBOL_GPL(nfsd_open_local_fh);
+
+/* Compile time type checking, not used by anything */
+static nfs_to_nfsd_open_local_fh_t __maybe_unused nfsd_open_local_fh_typecheck = nfsd_open_local_fh;
fs/nfsd/netns.h +7 -1
index 238fc4e56e53..e2d953f21dde 100644
@@ -13,6 +13,7 @@
 #include <linux/filelock.h>
 #include <linux/nfs4.h>
 #include <linux/percpu_counter.h>
+#include <linux/percpu-refcount.h>
 #include <linux/siphash.h>
 #include <linux/sunrpc/stats.h>
 
@@ -139,7 +140,9 @@ struct nfsd_net {
 
 	struct svc_info nfsd_info;
 #define nfsd_serv nfsd_info.serv
-
+	struct percpu_ref nfsd_serv_ref;
+	struct completion nfsd_serv_confirm_done;
+	struct completion nfsd_serv_free_done;
 
 	/*
 	 * clientid and stateid data for construction of net unique COPY
@@ -221,6 +224,9 @@ struct nfsd_net {
 extern bool nfsd_support_version(int vers);
 extern unsigned int nfsd_net_id;
 
+bool nfsd_serv_try_get(struct nfsd_net *nn);
+void nfsd_serv_put(struct nfsd_net *nn);
+
 void nfsd_copy_write_verifier(__be32 verf[2], struct nfsd_net *nn);
 void nfsd_reset_write_verifier(struct nfsd_net *nn);
 #endif /* __NFSD_NETNS_H__ */
fs/nfsd/nfssvc.c +14 -0
index b02eaaea7d62..c639fbe4d8c2 100644
@@ -192,6 +192,30 @@ int nfsd_minorversion(struct nfsd_net *nn, u32 minorversion, enum vers_op change
 	return 0;
 }
 
+bool nfsd_serv_try_get(struct nfsd_net *nn)
+{
+	return percpu_ref_tryget_live(&nn->nfsd_serv_ref);
+}
+
+void nfsd_serv_put(struct nfsd_net *nn)
+{
+	percpu_ref_put(&nn->nfsd_serv_ref);
+}
+
+static void nfsd_serv_done(struct percpu_ref *ref)
+{
+	struct nfsd_net *nn = container_of(ref, struct nfsd_net, nfsd_serv_ref);
+

fs/nfsd/nfssvc.c int nfsd_minorversion(struct nfsd_net *nn, u32 minorversion, enum vers_op change · line 218

Quoted 10 lines
fs/nfsd/nfssvc.c +10 -0
+	complete(&nn->nfsd_serv_confirm_done);
+}
+
+static void nfsd_serv_free(struct percpu_ref *ref)
+{
+	struct nfsd_net *nn = container_of(ref, struct nfsd_net, nfsd_serv_ref);
+
+	complete(&nn->nfsd_serv_free_done);
+}
+

Neil set a good precedent here: Perhaps nfsd_serv_try_get() and friends should be added as a separate patch with a patch description that explains their purpose and operation

Quoted 84 lines
fs/nfsd/nfssvc.c +15 -0
 /*
  * Maximum number of nfsd processes
  */
@@ -391,6 +415,7 @@ static void nfsd_shutdown_net(struct net *net)
 		lockd_down(net);
 		nn->lockd_up = false;
 	}
+	percpu_ref_exit(&nn->nfsd_serv_ref);
 	nn->nfsd_net_up = false;
 	nfsd_shutdown_generic();
 }
@@ -470,6 +495,13 @@ void nfsd_destroy_serv(struct net *net)
 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 	struct svc_serv *serv = nn->nfsd_serv;
 
+	lockdep_assert_held(&nfsd_mutex);
+
+	percpu_ref_kill_and_confirm(&nn->nfsd_serv_ref, nfsd_serv_done);
+	wait_for_completion(&nn->nfsd_serv_confirm_done);
+	wait_for_completion(&nn->nfsd_serv_free_done);
+	/* percpu_ref_exit is called in nfsd_shutdown_net */
+
 	spin_lock(&nfsd_notifier_lock);
 	nn->nfsd_serv = NULL;
 	spin_unlock(&nfsd_notifier_lock);
@@ -594,6 +626,13 @@ int nfsd_create_serv(struct net *net)
 	if (nn->nfsd_serv)
 		return 0;
 
+	error = percpu_ref_init(&nn->nfsd_serv_ref, nfsd_serv_free,
+				0, GFP_KERNEL);
+	if (error)
+		return error;
+	init_completion(&nn->nfsd_serv_free_done);
+	init_completion(&nn->nfsd_serv_confirm_done);
+
 	if (nfsd_max_blksize == 0)
 		nfsd_max_blksize = nfsd_get_default_max_blksize();
 	nfsd_reset_versions(nn);
fs/nfsd/trace.h +2 -1
index 77bbd23aa150..9c0610fdd11c 100644
@@ -86,7 +86,8 @@ DEFINE_NFSD_XDR_ERR_EVENT(cant_encode);
 		{ NFSD_MAY_NOT_BREAK_LEASE,	"NOT_BREAK_LEASE" },	\
 		{ NFSD_MAY_BYPASS_GSS,		"BYPASS_GSS" },		\
 		{ NFSD_MAY_READ_IF_EXEC,	"READ_IF_EXEC" },	\
-		{ NFSD_MAY_64BIT_COOKIE,	"64BIT_COOKIE" })
+		{ NFSD_MAY_64BIT_COOKIE,	"64BIT_COOKIE" },	\
+		{ NFSD_MAY_LOCALIO,		"LOCALIO" })
 
 TRACE_EVENT(nfsd_compound,
 	TP_PROTO(
fs/nfsd/vfs.h +10 -0
index 01947561d375..ec8a8aae540b 100644
@@ -33,6 +33,8 @@
 
 #define NFSD_MAY_64BIT_COOKIE		0x1000 /* 64 bit readdir cookies for >= NFSv3 */
 
+#define NFSD_MAY_LOCALIO		0x2000 /* for tracing, reflects when localio used */
+
 #define NFSD_MAY_CREATE		(NFSD_MAY_EXEC|NFSD_MAY_WRITE)
 #define NFSD_MAY_REMOVE		(NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC)
 
@@ -158,6 +160,14 @@ __be32		nfsd_permission(struct svc_cred *cred, struct svc_export *exp,
 
 void		nfsd_filp_close(struct file *fp);
 
+int		nfsd_open_local_fh(struct net *net,
+				   struct auth_domain *dom,
+				   struct rpc_clnt *rpc_clnt,
+				   const struct cred *cred,
+				   const struct nfs_fh *nfs_fh,
+				   const fmode_t fmode,
+				   struct nfsd_file **pnf);
+
 static inline int fh_want_write(struct svc_fh *fh)
 {
 	int ret;
Signature · 1 line
2.44.0
Signature · 1 line
Chuck Lever
#28 lore raw

Re: [PATCH v13 10/19] nfsd: add localio support

In reply to Mike Snitzer

fs/nfsd/localio.c line 75

Quoted 3 lines · Mike Snitzer
fs/nfsd/localio.c new file +3 -0
+
+	/* Save client creds before calling nfsd_file_acquire_local which calls nfsd_setuser */
+	save_cred = get_current_cred();

I don't think this belongs here. I would rather than nfsd_file_acquire_local() saved and restored the cred so it could be called without concern for internal implementation details.

Quoted 15 lines
fs/nfsd/localio.c new file +15 -0
+
+	/* nfs_fh -> svc_fh */
+	fh_init(&fh, NFS4_FHSIZE);
+	fh.fh_handle.fh_size = nfs_fh->size;
+	memcpy(fh.fh_handle.fh_raw, nfs_fh->data, nfs_fh->size);
+
+	if (fmode & FMODE_READ)
+		mayflags |= NFSD_MAY_READ;
+	if (fmode & FMODE_WRITE)
+		mayflags |= NFSD_MAY_WRITE;
+
+	rpcauth_map_clnt_to_svc_cred_local(rpc_clnt, cred, &rq_cred);
+
+	beres = nfsd_file_acquire_local(cl_nfssvc_net, &rq_cred, rpc_clnt->cl_vers,
+					cl_nfssvc_dom, &fh, mayflags, pnf);

fs/nfsd/localio.c line 100

Quoted 10 lines
fs/nfsd/localio.c new file +10 -0
+	if (beres) {
+		status = nfs_stat_to_errno(be32_to_cpu(beres));
+		goto out_fh_put;
+	}
+out_fh_put:
+	fh_put(&fh);
+	if (rq_cred.cr_group_info)
+		put_group_info(rq_cred.cr_group_info);
+	revert_creds(save_cred);
+	nfsd_serv_put(nn);

I think this is too early to be calling nfsd_serv_put(). I think it should be called when the IO completes - when nfs_to.nfsd_file_put() is called.

nfs_to.nfsd_open_local_fh() and nfs_to.nfsd_file_get() should each get a ref to the server.
nfsd_to.nfsd_file_put() should drop the ref.

Note that nfs_do.nfsd_file_get() would not exactly be nfsd_file_get. So maybe a different name would suit.

Thanks,
NeilBrown

#29 lore raw

Re: [PATCH v13 10/19] nfsd: add localio support

Quoted 8 lines · NeilBrown

fs/nfsd/localio.c line 75

Quoted 3 lines · Mike Snitzer
fs/nfsd/localio.c new file +3 -0
+
+	/* Save client creds before calling nfsd_file_acquire_local which calls nfsd_setuser */
+	save_cred = get_current_cred();

I don't think this belongs here. I would rather than nfsd_file_acquire_local() saved and restored the cred so it could be called without concern for internal implementation details.

OK, will fix.

Quoted 30 lines
Quoted 15 lines
fs/nfsd/localio.c new file +15 -0
+
+	/* nfs_fh -> svc_fh */
+	fh_init(&fh, NFS4_FHSIZE);
+	fh.fh_handle.fh_size = nfs_fh->size;
+	memcpy(fh.fh_handle.fh_raw, nfs_fh->data, nfs_fh->size);
+
+	if (fmode & FMODE_READ)
+		mayflags |= NFSD_MAY_READ;
+	if (fmode & FMODE_WRITE)
+		mayflags |= NFSD_MAY_WRITE;
+
+	rpcauth_map_clnt_to_svc_cred_local(rpc_clnt, cred, &rq_cred);
+
+	beres = nfsd_file_acquire_local(cl_nfssvc_net, &rq_cred, rpc_clnt->cl_vers,
+					cl_nfssvc_dom, &fh, mayflags, pnf);

fs/nfsd/localio.c line 100

Quoted 10 lines
fs/nfsd/localio.c new file +10 -0
+	if (beres) {
+		status = nfs_stat_to_errno(be32_to_cpu(beres));
+		goto out_fh_put;
+	}
+out_fh_put:
+	fh_put(&fh);
+	if (rq_cred.cr_group_info)
+		put_group_info(rq_cred.cr_group_info);
+	revert_creds(save_cred);
+	nfsd_serv_put(nn);

I think this is too early to be calling nfsd_serv_put(). I think it should be called when the IO completes - when nfs_to.nfsd_file_put() is called.

Quoted 6 lines

nfs_to.nfsd_open_local_fh() and nfs_to.nfsd_file_get() should each get a ref to the server.
nfsd_to.nfsd_file_put() should drop the ref.

Note that nfs_do.nfsd_file_get() would not exactly be nfsd_file_get. So maybe a different name would suit.

Yes, you're right. I'll get it fixed.

Thanks!

#30 lore raw

PATCH v13 11/19 nfsd: implement server support for NFS_LOCALIO_PROGRAM

In reply to Mike Snitzer

The LOCALIO auxiliary RPC protocol consists of a single "UUID_IS_LOCAL" RPC method that allows the Linux NFS client to verify the local Linux NFS server can see the nonce (single-use UUID) the client generated and made available in nfs_common. This protocol isn't part of an IETF standard, nor does it need to be considering it is Linux-to-Linux auxiliary RPC protocol that amounts to an implementation detail.

The UUID_IS_LOCAL method encodes the client generated uuid_t in terms of the fixed UUID_SIZE (16 bytes). The fixed size opaque encode and decode XDR methods are used instead of the less efficient variable sized methods.

The RPC program number for the NFS_LOCALIO_PROGRAM is 400122 (as assigned by IANA, see https://www.iana.org/assignments/rpc-program-numbers/ ):
Linux Kernel Organization 400122 nfslocalio

[neilb: factored out and simplified single localio protocol]
---
 fs/nfsd/localio.c   | 75 +++++++++++++++++++++++++++++++++++++++++++++
 fs/nfsd/nfsd.h      |  4 +++
 fs/nfsd/nfssvc.c    | 23 +++++++++++++-
 include/linux/nfs.h |  7 +++++
 4 files changed, 108 insertions(+), 1 deletion(-)
fs/nfsd/localio.c +75 -0
index 8e0cce835f81..2ceab49f3cb6 100644
@@ -13,12 +13,15 @@
 #include <linux/nfs.h>
 #include <linux/nfs_common.h>
 #include <linux/nfslocalio.h>
+#include <linux/nfs_fs.h>
+#include <linux/nfs_xdr.h>
 #include <linux/string.h>
 
 #include "nfsd.h"
 #include "vfs.h"
 #include "netns.h"
 #include "filecache.h"
+#include "cache.h"
 
 /**
  * nfsd_open_local_fh - lookup a local filehandle @nfs_fh and map to @file
@@ -106,3 +109,75 @@ EXPORT_SYMBOL_GPL(nfsd_open_local_fh);
 
 /* Compile time type checking, not used by anything */
 static nfs_to_nfsd_open_local_fh_t __maybe_unused nfsd_open_local_fh_typecheck = nfsd_open_local_fh;
+
+/*
+ * UUID_IS_LOCAL XDR functions
+ */
+
+static __be32 localio_proc_null(struct svc_rqst *rqstp)
+{
+	return rpc_success;
+}
+
+struct localio_uuidarg {
+	uuid_t			uuid;
+};
+
+static __be32 localio_proc_uuid_is_local(struct svc_rqst *rqstp)
+{
+	struct localio_uuidarg *argp = rqstp->rq_argp;
+
+	(void) nfs_uuid_is_local(&argp->uuid, SVC_NET(rqstp),
+				 rqstp->rq_client);
+
+	return rpc_success;
+}
+
+static bool localio_decode_uuidarg(struct svc_rqst *rqstp,
+				   struct xdr_stream *xdr)
+{
+	struct localio_uuidarg *argp = rqstp->rq_argp;
+	u8 uuid[UUID_SIZE];
+
+	if (decode_opaque_fixed(xdr, uuid, UUID_SIZE))
+		return false;
+	import_uuid(&argp->uuid, uuid);
+
+	return true;
+}
+
+static const struct svc_procedure localio_procedures1[] = {
+	[LOCALIOPROC_NULL] = {
+		.pc_func = localio_proc_null,
+		.pc_decode = nfssvc_decode_voidarg,
+		.pc_encode = nfssvc_encode_voidres,
+		.pc_argsize = sizeof(struct nfsd_voidargs),
+		.pc_ressize = sizeof(struct nfsd_voidres),
+		.pc_cachetype = RC_NOCACHE,
+		.pc_xdrressize = 0,
+		.pc_name = "NULL",
+	},
+	[LOCALIOPROC_UUID_IS_LOCAL] = {
+		.pc_func = localio_proc_uuid_is_local,
+		.pc_decode = localio_decode_uuidarg,
+		.pc_encode = nfssvc_encode_voidres,
+		.pc_argsize = sizeof(struct localio_uuidarg),
+		.pc_argzero = sizeof(struct localio_uuidarg),
+		.pc_ressize = sizeof(struct nfsd_voidres),
+		.pc_cachetype = RC_NOCACHE,
+		.pc_name = "UUID_IS_LOCAL",
+	},
+};
+
+#define LOCALIO_NR_PROCEDURES ARRAY_SIZE(localio_procedures1)
+static DEFINE_PER_CPU_ALIGNED(unsigned long,
+			      localio_count[LOCALIO_NR_PROCEDURES]);
+const struct svc_version localio_version1 = {
+	.vs_vers	= 1,
+	.vs_nproc	= LOCALIO_NR_PROCEDURES,
+	.vs_proc	= localio_procedures1,
+	.vs_dispatch	= nfsd_dispatch,
+	.vs_count	= localio_count,
+	.vs_xdrsize	= XDR_QUADLEN(UUID_SIZE),
+	.vs_hidden	= true,
+};
fs/nfsd/nfsd.h +4 -0
index b0d3e82d6dcd..232a873dc53a 100644
@@ -146,6 +146,10 @@ extern const struct svc_version nfsd_acl_version3;
 #endif
 #endif
 
+#if IS_ENABLED(CONFIG_NFSD_LOCALIO)
+extern const struct svc_version localio_version1;
+#endif
+
 struct nfsd_net;
 
 enum vers_op {NFSD_SET, NFSD_CLEAR, NFSD_TEST, NFSD_AVAIL };
fs/nfsd/nfssvc.c +22 -1
index c639fbe4d8c2..5f8680ab1013 100644
@@ -79,6 +79,15 @@ DEFINE_SPINLOCK(nfsd_drc_lock);
 unsigned long	nfsd_drc_max_mem;
 unsigned long	nfsd_drc_mem_used;
 
+#if IS_ENABLED(CONFIG_NFSD_LOCALIO)
+static const struct svc_version *localio_versions[] = {
+	[1] = &localio_version1,
+};
+
+#define NFSD_LOCALIO_NRVERS		ARRAY_SIZE(localio_versions)
+
+#endif /* CONFIG_NFSD_LOCALIO */
+
 #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
 static const struct svc_version *nfsd_acl_version[] = {
 # if defined(CONFIG_NFSD_V2_ACL)
@@ -127,6 +136,18 @@ struct svc_program		nfsd_programs[] = {
 	.pg_rpcbind_set		= nfsd_acl_rpcbind_set,
 	},
 #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */
+#if IS_ENABLED(CONFIG_NFSD_LOCALIO)
+	{
+	.pg_prog		= NFS_LOCALIO_PROGRAM,
+	.pg_nvers		= NFSD_LOCALIO_NRVERS,
+	.pg_vers		= localio_versions,
+	.pg_name		= "nfslocalio",
+	.pg_class		= "nfsd",
+	.pg_authenticate	= svc_set_client,
+	.pg_init_request	= svc_generic_init_request,
+	.pg_rpcbind_set		= svc_generic_rpcbind_set,
+	}
+#endif /* IS_ENABLED(CONFIG_NFSD_LOCALIO) */
 };
 
 bool nfsd_support_version(int vers)
@@ -944,7 +965,7 @@ nfsd(void *vrqstp)
 }
 
 /**
- * nfsd_dispatch - Process an NFS or NFSACL Request
+ * nfsd_dispatch - Process an NFS or NFSACL or LOCALIO Request
  * @rqstp: incoming request
  *
  * This RPC dispatcher integrates the NFS server's duplicate reply cache.
include/linux/nfs.h +7 -0
index ceb70a926b95..5ff1a5b3b00c 100644
@@ -13,6 +13,13 @@
 #include <linux/crc32.h>
 #include <uapi/linux/nfs.h>
 
+/* The localio program is entirely private to Linux and is
+ * NOT part of the uapi.
+ */
+#define NFS_LOCALIO_PROGRAM		400122
+#define LOCALIOPROC_NULL		0
+#define LOCALIOPROC_UUID_IS_LOCAL	1
+
 /*
  * This is the kernel NFS client file handle representation
  */
Signature · 1 line
2.44.0
Acked-by on v13 11/19
#31 lore raw

Re: [PATCH v13 11/19] nfsd: implement server support for NFS_LOCALIO_PROGRAM

Quoted 7 lines · Mike Snitzer

The LOCALIO auxiliary RPC protocol consists of a single "UUID_IS_LOCAL" RPC method that allows the Linux NFS client to verify the local Linux NFS server can see the nonce (single-use UUID) the client generated and made available in nfs_common. This protocol isn't part of an IETF standard, nor does it need to be considering it is Linux-to-Linux auxiliary RPC protocol that amounts to an implementation detail.

Quoted 8 lines

The UUID_IS_LOCAL method encodes the client generated uuid_t in terms of the fixed UUID_SIZE (16 bytes). The fixed size opaque encode and decode XDR methods are used instead of the less efficient variable sized methods.

The RPC program number for the NFS_LOCALIO_PROGRAM is 400122 (as assigned by IANA, see https://www.iana.org/assignments/rpc-program-numbers/ ):
Linux Kernel Organization 400122 nfslocalio

Since it is a major design point for this patch, the patch description needs to mention that the server expects LOCALIO RPCs on the same transport as NFS and NFSACL operations.

Quoted 191 lines
[neilb: factored out and simplified single localio protocol]
---
 fs/nfsd/localio.c   | 75 +++++++++++++++++++++++++++++++++++++++++++++
 fs/nfsd/nfsd.h      |  4 +++
 fs/nfsd/nfssvc.c    | 23 +++++++++++++-
 include/linux/nfs.h |  7 +++++
 4 files changed, 108 insertions(+), 1 deletion(-)
fs/nfsd/localio.c +75 -0
index 8e0cce835f81..2ceab49f3cb6 100644
@@ -13,12 +13,15 @@
 #include <linux/nfs.h>
 #include <linux/nfs_common.h>
 #include <linux/nfslocalio.h>
+#include <linux/nfs_fs.h>
+#include <linux/nfs_xdr.h>
 #include <linux/string.h>
 
 #include "nfsd.h"
 #include "vfs.h"
 #include "netns.h"
 #include "filecache.h"
+#include "cache.h"
 
 /**
  * nfsd_open_local_fh - lookup a local filehandle @nfs_fh and map to @file
@@ -106,3 +109,75 @@ EXPORT_SYMBOL_GPL(nfsd_open_local_fh);
 
 /* Compile time type checking, not used by anything */
 static nfs_to_nfsd_open_local_fh_t __maybe_unused nfsd_open_local_fh_typecheck = nfsd_open_local_fh;
+
+/*
+ * UUID_IS_LOCAL XDR functions
+ */
+
+static __be32 localio_proc_null(struct svc_rqst *rqstp)
+{
+	return rpc_success;
+}
+
+struct localio_uuidarg {
+	uuid_t			uuid;
+};
+
+static __be32 localio_proc_uuid_is_local(struct svc_rqst *rqstp)
+{
+	struct localio_uuidarg *argp = rqstp->rq_argp;
+
+	(void) nfs_uuid_is_local(&argp->uuid, SVC_NET(rqstp),
+				 rqstp->rq_client);
+
+	return rpc_success;
+}
+
+static bool localio_decode_uuidarg(struct svc_rqst *rqstp,
+				   struct xdr_stream *xdr)
+{
+	struct localio_uuidarg *argp = rqstp->rq_argp;
+	u8 uuid[UUID_SIZE];
+
+	if (decode_opaque_fixed(xdr, uuid, UUID_SIZE))
+		return false;
+	import_uuid(&argp->uuid, uuid);
+
+	return true;
+}
+
+static const struct svc_procedure localio_procedures1[] = {
+	[LOCALIOPROC_NULL] = {
+		.pc_func = localio_proc_null,
+		.pc_decode = nfssvc_decode_voidarg,
+		.pc_encode = nfssvc_encode_voidres,
+		.pc_argsize = sizeof(struct nfsd_voidargs),
+		.pc_ressize = sizeof(struct nfsd_voidres),
+		.pc_cachetype = RC_NOCACHE,
+		.pc_xdrressize = 0,
+		.pc_name = "NULL",
+	},
+	[LOCALIOPROC_UUID_IS_LOCAL] = {
+		.pc_func = localio_proc_uuid_is_local,
+		.pc_decode = localio_decode_uuidarg,
+		.pc_encode = nfssvc_encode_voidres,
+		.pc_argsize = sizeof(struct localio_uuidarg),
+		.pc_argzero = sizeof(struct localio_uuidarg),
+		.pc_ressize = sizeof(struct nfsd_voidres),
+		.pc_cachetype = RC_NOCACHE,
+		.pc_name = "UUID_IS_LOCAL",
+	},
+};
+
+#define LOCALIO_NR_PROCEDURES ARRAY_SIZE(localio_procedures1)
+static DEFINE_PER_CPU_ALIGNED(unsigned long,
+			      localio_count[LOCALIO_NR_PROCEDURES]);
+const struct svc_version localio_version1 = {
+	.vs_vers	= 1,
+	.vs_nproc	= LOCALIO_NR_PROCEDURES,
+	.vs_proc	= localio_procedures1,
+	.vs_dispatch	= nfsd_dispatch,
+	.vs_count	= localio_count,
+	.vs_xdrsize	= XDR_QUADLEN(UUID_SIZE),
+	.vs_hidden	= true,
+};
fs/nfsd/nfsd.h +4 -0
index b0d3e82d6dcd..232a873dc53a 100644
@@ -146,6 +146,10 @@ extern const struct svc_version nfsd_acl_version3;
 #endif
 #endif
 
+#if IS_ENABLED(CONFIG_NFSD_LOCALIO)
+extern const struct svc_version localio_version1;
+#endif
+
 struct nfsd_net;
 
 enum vers_op {NFSD_SET, NFSD_CLEAR, NFSD_TEST, NFSD_AVAIL };
fs/nfsd/nfssvc.c +22 -1
index c639fbe4d8c2..5f8680ab1013 100644
@@ -79,6 +79,15 @@ DEFINE_SPINLOCK(nfsd_drc_lock);
 unsigned long	nfsd_drc_max_mem;
 unsigned long	nfsd_drc_mem_used;
 
+#if IS_ENABLED(CONFIG_NFSD_LOCALIO)
+static const struct svc_version *localio_versions[] = {
+	[1] = &localio_version1,
+};
+
+#define NFSD_LOCALIO_NRVERS		ARRAY_SIZE(localio_versions)
+
+#endif /* CONFIG_NFSD_LOCALIO */
+
 #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
 static const struct svc_version *nfsd_acl_version[] = {
 # if defined(CONFIG_NFSD_V2_ACL)
@@ -127,6 +136,18 @@ struct svc_program		nfsd_programs[] = {
 	.pg_rpcbind_set		= nfsd_acl_rpcbind_set,
 	},
 #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */
+#if IS_ENABLED(CONFIG_NFSD_LOCALIO)
+	{
+	.pg_prog		= NFS_LOCALIO_PROGRAM,
+	.pg_nvers		= NFSD_LOCALIO_NRVERS,
+	.pg_vers		= localio_versions,
+	.pg_name		= "nfslocalio",
+	.pg_class		= "nfsd",
+	.pg_authenticate	= svc_set_client,
+	.pg_init_request	= svc_generic_init_request,
+	.pg_rpcbind_set		= svc_generic_rpcbind_set,
+	}
+#endif /* IS_ENABLED(CONFIG_NFSD_LOCALIO) */
 };
 
 bool nfsd_support_version(int vers)
@@ -944,7 +965,7 @@ nfsd(void *vrqstp)
 }
 
 /**
- * nfsd_dispatch - Process an NFS or NFSACL Request
+ * nfsd_dispatch - Process an NFS or NFSACL or LOCALIO Request
  * @rqstp: incoming request
  *
  * This RPC dispatcher integrates the NFS server's duplicate reply cache.
include/linux/nfs.h +7 -0
index ceb70a926b95..5ff1a5b3b00c 100644
@@ -13,6 +13,13 @@
 #include <linux/crc32.h>
 #include <uapi/linux/nfs.h>
 
+/* The localio program is entirely private to Linux and is
+ * NOT part of the uapi.
+ */
+#define NFS_LOCALIO_PROGRAM		400122
+#define LOCALIOPROC_NULL		0
+#define LOCALIOPROC_UUID_IS_LOCAL	1
+
 /*
  * This is the kernel NFS client file handle representation
  */
Signature · 1 line
2.44.0
Signature · 1 line
Chuck Lever
#32 lore raw

PATCH v13 12/19 nfs: pass struct nfsd_file to nfs_init_pgio and nfs_init_commit

In reply to Mike Snitzer

The localio nfsd_file will be passed, in future commits, by callers that enable localio support (for both regular NFS and pNFS IO).

[Derived from patch authored by Weston Andros Adamson, but switched
from passing struct file to struct nfsd_file]

---
 fs/nfs/filelayout/filelayout.c         | 6 +++---
 fs/nfs/flexfilelayout/flexfilelayout.c | 6 +++---
 fs/nfs/internal.h                      | 6 ++++--
 fs/nfs/pagelist.c                      | 6 ++++--
 fs/nfs/pnfs_nfs.c                      | 2 +-
 fs/nfs/write.c                         | 5 +++--
 6 files changed, 18 insertions(+), 13 deletions(-)
fs/nfs/filelayout/filelayout.c +3 -3
index b6e9aeaf4ce2..d39a1f58e18d 100644
@@ -488,7 +488,7 @@ filelayout_read_pagelist(struct nfs_pgio_header *hdr)
 	/* Perform an asynchronous read to ds */
 	nfs_initiate_pgio(ds_clnt, hdr, hdr->cred,
 			  NFS_PROTO(hdr->inode), &filelayout_read_call_ops,
-			  0, RPC_TASK_SOFTCONN);
+			  0, RPC_TASK_SOFTCONN, NULL);
 	return PNFS_ATTEMPTED;
 }
 
@@ -530,7 +530,7 @@ filelayout_write_pagelist(struct nfs_pgio_header *hdr, int sync)
 	/* Perform an asynchronous write */
 	nfs_initiate_pgio(ds_clnt, hdr, hdr->cred,
 			  NFS_PROTO(hdr->inode), &filelayout_write_call_ops,
-			  sync, RPC_TASK_SOFTCONN);
+			  sync, RPC_TASK_SOFTCONN, NULL);
 	return PNFS_ATTEMPTED;
 }
 
@@ -1011,7 +1011,7 @@ static int filelayout_initiate_commit(struct nfs_commit_data *data, int how)
 		data->args.fh = fh;
 	return nfs_initiate_commit(ds_clnt, data, NFS_PROTO(data->inode),
 				   &filelayout_commit_call_ops, how,
-				   RPC_TASK_SOFTCONN);
+				   RPC_TASK_SOFTCONN, NULL);
 out_err:
 	pnfs_generic_prepare_to_resend_writes(data);
 	pnfs_generic_commit_release(data);
fs/nfs/flexfilelayout/flexfilelayout.c +3 -3
index d4d551ffea7b..01ee52551a63 100644
@@ -1806,7 +1806,7 @@ ff_layout_read_pagelist(struct nfs_pgio_header *hdr)
 	nfs_initiate_pgio(ds_clnt, hdr, ds_cred, ds->ds_clp->rpc_ops,
 			  vers == 3 ? &ff_layout_read_call_ops_v3 :
 				      &ff_layout_read_call_ops_v4,
-			  0, RPC_TASK_SOFTCONN);
+			  0, RPC_TASK_SOFTCONN, NULL);
 	put_cred(ds_cred);
 	return PNFS_ATTEMPTED;
 
@@ -1874,7 +1874,7 @@ ff_layout_write_pagelist(struct nfs_pgio_header *hdr, int sync)
 	nfs_initiate_pgio(ds_clnt, hdr, ds_cred, ds->ds_clp->rpc_ops,
 			  vers == 3 ? &ff_layout_write_call_ops_v3 :
 				      &ff_layout_write_call_ops_v4,
-			  sync, RPC_TASK_SOFTCONN);
+			  sync, RPC_TASK_SOFTCONN, NULL);
 	put_cred(ds_cred);
 	return PNFS_ATTEMPTED;
 
@@ -1949,7 +1949,7 @@ static int ff_layout_initiate_commit(struct nfs_commit_data *data, int how)
 	ret = nfs_initiate_commit(ds_clnt, data, ds->ds_clp->rpc_ops,
 				   vers == 3 ? &ff_layout_commit_call_ops_v3 :
 					       &ff_layout_commit_call_ops_v4,
-				   how, RPC_TASK_SOFTCONN);
+				   how, RPC_TASK_SOFTCONN, NULL);
 	put_cred(ds_cred);
 	return ret;
 out_err:
fs/nfs/internal.h +4 -2
index 5902a9beca1f..eed35dcff54f 100644
@@ -308,7 +308,8 @@ void nfs_pgio_header_free(struct nfs_pgio_header *);
 int nfs_generic_pgio(struct nfs_pageio_descriptor *, struct nfs_pgio_header *);
 int nfs_initiate_pgio(struct rpc_clnt *clnt, struct nfs_pgio_header *hdr,
 		      const struct cred *cred, const struct nfs_rpc_ops *rpc_ops,
-		      const struct rpc_call_ops *call_ops, int how, int flags);
+		      const struct rpc_call_ops *call_ops, int how, int flags,
+		      struct nfsd_file *localio);
 void nfs_free_request(struct nfs_page *req);
 struct nfs_pgio_mirror *
 nfs_pgio_current_mirror(struct nfs_pageio_descriptor *desc);
@@ -528,7 +529,8 @@ extern int nfs_initiate_commit(struct rpc_clnt *clnt,
 			       struct nfs_commit_data *data,
 			       const struct nfs_rpc_ops *nfs_ops,
 			       const struct rpc_call_ops *call_ops,
-			       int how, int flags);
+			       int how, int flags,
+			       struct nfsd_file *localio);
 extern void nfs_init_commit(struct nfs_commit_data *data,
 			    struct list_head *head,
 			    struct pnfs_layout_segment *lseg,
fs/nfs/pagelist.c +4 -2
index 04124f226665..50f3d6c9ac2a 100644
@@ -731,7 +731,8 @@ static void nfs_pgio_prepare(struct rpc_task *task, void *calldata)
 
 int nfs_initiate_pgio(struct rpc_clnt *clnt, struct nfs_pgio_header *hdr,
 		      const struct cred *cred, const struct nfs_rpc_ops *rpc_ops,
-		      const struct rpc_call_ops *call_ops, int how, int flags)
+		      const struct rpc_call_ops *call_ops, int how, int flags,
+		      struct nfsd_file *localio)
 {
 	struct rpc_task *task;
 	struct rpc_message msg = {
@@ -961,7 +962,8 @@ static int nfs_generic_pg_pgios(struct nfs_pageio_descriptor *desc)
 					NFS_PROTO(hdr->inode),
 					desc->pg_rpc_callops,
 					desc->pg_ioflags,
-					RPC_TASK_CRED_NOREF | task_flags);
+					RPC_TASK_CRED_NOREF | task_flags,
+					NULL);
 	}
 	return ret;
 }
fs/nfs/pnfs_nfs.c +1 -1
index a74ee69a2fa6..dbef837e871a 100644
@@ -490,7 +490,7 @@ pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages,
 			nfs_initiate_commit(NFS_CLIENT(inode), data,
 					    NFS_PROTO(data->inode),
 					    data->mds_ops, how,
-					    RPC_TASK_CRED_NOREF);
+					    RPC_TASK_CRED_NOREF, NULL);
 		} else {
 			nfs_init_commit(data, NULL, data->lseg, cinfo);
 			initiate_commit(data, how);
fs/nfs/write.c +3 -2
index d074d0ceb4f0..04d0b5b95f4f 100644
@@ -1663,7 +1663,8 @@ EXPORT_SYMBOL_GPL(nfs_commitdata_release);
 int nfs_initiate_commit(struct rpc_clnt *clnt, struct nfs_commit_data *data,
 			const struct nfs_rpc_ops *nfs_ops,
 			const struct rpc_call_ops *call_ops,
-			int how, int flags)
+			int how, int flags,
+			struct nfsd_file *localio)
 {
 	struct rpc_task *task;
 	int priority = flush_task_priority(how);
@@ -1809,7 +1810,7 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how,
 		task_flags = RPC_TASK_MOVEABLE;
 	return nfs_initiate_commit(NFS_CLIENT(inode), data, NFS_PROTO(inode),
 				   data->mds_ops, how,
-				   RPC_TASK_CRED_NOREF | task_flags);
+				   RPC_TASK_CRED_NOREF | task_flags, NULL);
 }
 
 /*
Signature · 1 line
2.44.0
#33 lore raw

PATCH v13 13/19 nfs: add localio support

In reply to Mike Snitzer

From: Weston Andros Adamson <[email protected]>

Add client support for bypassing NFS for localhost reads, writes, and commits. This is only useful when the client and the server are running on the same host.

nfs_local_probe() is stubbed out, later commits will enable client and server handshake via a Linux-only LOCALIO auxiliary RPC protocol.

This has dynamic binding with the nfsd module (via nfs_localio module which is part of nfs_common). Localio will only work if nfsd is already loaded.

The "localio_enabled" nfs kernel module parameter can be used to disable and enable the ability to use localio support.

CONFIG_NFS_LOCALIO controls the client enablement.

Lastly, localio uses an nfsd_file to initiate all IO. To make proper use of nfsd_file (and nfsd's filecache) its lifetime (duration before nfsd_file_put is called) must extend until after commit, read and write operations. So rather than immediately call nfsd_file_put() in nfs_local_open_fh(), nfsd_file_put() isn't called until nfs_local_pgio_release() for read/write and not until nfs_local_release_commit_data() for commit.

---
 fs/nfs/Kconfig            |  14 +
 fs/nfs/Makefile           |   1 +
 fs/nfs/client.c           |  11 +
 fs/nfs/internal.h         |  54 ++++
 fs/nfs/localio.c          | 618 ++++++++++++++++++++++++++++++++++++++
 fs/nfs/nfstrace.h         |  61 ++++
 fs/nfs/pagelist.c         |   4 +
 fs/nfs/write.c            |   3 +
 include/linux/nfs.h       |   2 +
 include/linux/nfs_fs_sb.h |  10 +
 10 files changed, 778 insertions(+)
 create mode 100644 fs/nfs/localio.c
fs/nfs/Kconfig +14 -0
index 0eb20012792f..d52a1df28f69 100644
@@ -87,6 +87,20 @@ config NFS_V4
 
 	  If unsure, say Y.
 
+config NFS_LOCALIO
+	bool "NFS client support for the LOCALIO auxiliary protocol"
+	depends on NFS_FS
+	select NFS_COMMON_LOCALIO_SUPPORT
+	help
+	  Some NFS servers support an auxiliary NFS LOCALIO protocol
+	  that is not an official part of the NFS protocol.
+
+	  This option enables support for the LOCALIO protocol in the
+	  kernel's NFS client.  Enable this to bypass using the NFS
+	  protocol when issuing reads, writes and commits to the server.
+
+	  If unsure, say N.
+
 config NFS_SWAP
 	bool "Provide swap over NFS support"
 	default n
fs/nfs/Makefile +1 -0
index 5f6db37f461e..9fb2f2cac87e 100644
@@ -13,6 +13,7 @@ nfs-y 			:= client.o dir.o file.o getroot.o inode.o super.o \
 nfs-$(CONFIG_ROOT_NFS)	+= nfsroot.o
 nfs-$(CONFIG_SYSCTL)	+= sysctl.o
 nfs-$(CONFIG_NFS_FSCACHE) += fscache.o
+nfs-$(CONFIG_NFS_LOCALIO) += localio.o
 
 obj-$(CONFIG_NFS_V2) += nfsv2.o
 nfsv2-y := nfs2super.o proc.o nfs2xdr.o
fs/nfs/client.c +11 -0
index 8286edd6062d..95e8d76bd49f 100644
@@ -178,6 +178,14 @@ struct nfs_client *nfs_alloc_client(const struct nfs_client_initdata *cl_init)
 	clp->cl_max_connect = cl_init->max_connect ? cl_init->max_connect : 1;
 	clp->cl_net = get_net(cl_init->net);
 
+#if IS_ENABLED(CONFIG_NFS_LOCALIO)
+	seqlock_init(&clp->cl_boot_lock);
+	ktime_get_real_ts64(&clp->cl_nfssvc_boot);
+	clp->cl_rpcclient_localio = ERR_PTR(-EINVAL);
+	clp->cl_nfssvc_net = NULL;
+	clp->cl_nfssvc_dom = NULL;
+#endif /* CONFIG_NFS_LOCALIO */
+
 	clp->cl_principal = "*";
 	clp->cl_xprtsec = cl_init->xprtsec;
 	return clp;
@@ -233,6 +241,8 @@ static void pnfs_init_server(struct nfs_server *server)
  */
 void nfs_free_client(struct nfs_client *clp)
 {
+	nfs_local_disable(clp);
+
 	/* -EIO all pending I/O */
 	if (!IS_ERR(clp->cl_rpcclient))
 		rpc_shutdown_client(clp->cl_rpcclient);
@@ -424,6 +434,7 @@ struct nfs_client *nfs_get_client(const struct nfs_client_initdata *cl_init)
 			list_add_tail(&new->cl_share_link,
 					&nn->nfs_client_list);
 			spin_unlock(&nn->nfs_client_lock);
+			nfs_local_probe(new);
 			return rpc_ops->init_client(new, cl_init);
 		}
 
fs/nfs/internal.h +54 -0

60 lines not on this page · raw message

fs/nfs/localio.c new file +618 -0

618 lines not on this page · raw message

fs/nfs/nfstrace.h +61 -0

67 lines not on this page · raw message

fs/nfs/pagelist.c +4 -0
index 50f3d6c9ac2a..97d5524c379a 100644
@@ -762,6 +762,10 @@ int nfs_initiate_pgio(struct rpc_clnt *clnt, struct nfs_pgio_header *hdr,
 		hdr->args.count,
 		(unsigned long long)hdr->args.offset);
 
+	if (localio)
+		return nfs_local_doio(NFS_SERVER(hdr->inode)->nfs_client,
+				      localio, hdr, call_ops);
+
 	task = rpc_run_task(&task_setup_data);
 	if (IS_ERR(task))
 		return PTR_ERR(task);
fs/nfs/write.c +3 -0
index 04d0b5b95f4f..404cc5281e6a 100644
@@ -1693,6 +1693,9 @@ int nfs_initiate_commit(struct rpc_clnt *clnt, struct nfs_commit_data *data,
 
 	dprintk("NFS: initiated commit call\n");
 
+	if (localio)
+		return nfs_local_commit(localio, data, call_ops, how);
+
 	task = rpc_run_task(&task_setup_data);
 	if (IS_ERR(task))
 		return PTR_ERR(task);
include/linux/nfs.h +2 -0
index 5ff1a5b3b00c..89ef8c5e98db 100644
@@ -8,6 +8,8 @@
 #ifndef _LINUX_NFS_H
 #define _LINUX_NFS_H
 
+#include <linux/cred.h>
+#include <linux/sunrpc/auth.h>
 #include <linux/sunrpc/msg_prot.h>
 #include <linux/string.h>
 #include <linux/crc32.h>
include/linux/nfs_fs_sb.h +10 -0
index 1df86ab98c77..10453c6f8ca8 100644
@@ -8,6 +8,7 @@
 #include <linux/wait.h>
 #include <linux/nfs_xdr.h>
 #include <linux/sunrpc/xprt.h>
+#include <linux/nfslocalio.h>
 
 #include <linux/atomic.h>
 #include <linux/refcount.h>
@@ -49,6 +50,7 @@ struct nfs_client {
 #define NFS_CS_DS		7		/* - Server is a DS */
 #define NFS_CS_REUSEPORT	8		/* - reuse src port on reconnect */
 #define NFS_CS_PNFS		9		/* - Server used for pnfs */
+#define NFS_CS_LOCAL_IO		10		/* - client is local */
 	struct sockaddr_storage	cl_addr;	/* server identifier */
 	size_t			cl_addrlen;
 	char *			cl_hostname;	/* hostname of server */
@@ -125,6 +127,14 @@ struct nfs_client {
 	struct net		*cl_net;
 	struct list_head	pending_cb_stateids;
 	struct rcu_head		rcu;
+
+#if IS_ENABLED(CONFIG_NFS_LOCALIO)
+	struct timespec64	cl_nfssvc_boot;
+	seqlock_t		cl_boot_lock;
+	struct rpc_clnt *	cl_rpcclient_localio;
+	struct net *	        cl_nfssvc_net;
+	struct auth_domain *	cl_nfssvc_dom;
+#endif /* CONFIG_NFS_LOCALIO */
 };
 
 /*
Signature · 1 line
2.44.0
#34 lore raw

Re: [PATCH v13 13/19] nfs: add localio support

fs/nfs/localio.c line 79

Quoted 6 lines · Mike Snitzer
fs/nfs/localio.c new file +6 -0
+/*
+ * nfs_local_disable - disable local i/o for an nfs_client
+ */
+void nfs_local_disable(struct nfs_client *clp)
+{
+	if (test_and_clear_bit(NFS_CS_LOCAL_IO, &clp->cl_flags)) {

This use of test_and_clear_bit() led me to think about the atomicity requirements here.

If this were called (on a different CPU) between the nfs_server_is_local() test in nfs_local_file_open() and the nfs_local_open_fh() call which follows, then nfs_to.nfsd_open_local_fh() could be passed a NULL client which wouldn't end well.

Maybe we need an rcu-protected auth_domain_tryget() in nfsd_local_open_fh(), and the auth_domain_put() below need to be behind call_rcu().

Or something...

Quoted 16 lines
fs/nfs/localio.c new file +14 -0
+		trace_nfs_local_disable(clp);
+		clp->cl_nfssvc_net = NULL;
+		if (clp->cl_nfssvc_dom) {
+			auth_domain_put(clp->cl_nfssvc_dom);
+			clp->cl_nfssvc_dom = NULL;
+		}
+	}
+}
 
 
+
+struct nfsd_file *
+nfs_local_file_open(struct nfs_client *clp, const struct cred *cred,
+		    struct nfs_fh *fh, struct nfs_open_context *ctx)
+{
+	struct nfsd_file *nf;

fs/nfs/localio.c line 144

Quoted 10 lines
fs/nfs/localio.c new file +10 -0
+
+	if (!nfs_server_is_local(clp))
+		return NULL;
+
+	nf = nfs_local_open_fh(clp, cred, fh, ctx->mode);
+	if (IS_ERR(nf))
+		return NULL;
+
+	return nf;
+}

It isn't clear to be what nfs_local_file_open() adds any value over nfs_local_open_fh. The nfs_server_is_local() test is cheap and can safely be in nfs_local_open_fh(), and the current callers of nfs_local_file_open() can simple pass ctx->mode to _fh instead of passing ctx to _open.

Thanks,
NeilBrown

#35 lore raw

PATCH v13 14/19 nfs: enable localio for non-pNFS IO

In reply to Mike Snitzer

From: Trond Myklebust <[email protected]>

Try a local open of the file being written to, and if it succeeds, then use localio to issue IO.

---
 fs/nfs/pagelist.c | 8 +++++++-
 fs/nfs/write.c    | 6 +++++-
 2 files changed, 12 insertions(+), 2 deletions(-)
fs/nfs/pagelist.c +7 -1
index 97d5524c379a..6f836b66ef79 100644
@@ -958,6 +958,12 @@ static int nfs_generic_pg_pgios(struct nfs_pageio_descriptor *desc)
 	nfs_pgheader_init(desc, hdr, nfs_pgio_header_free);
 	ret = nfs_generic_pgio(desc, hdr);
 	if (ret == 0) {
+		struct nfs_client *clp = NFS_SERVER(hdr->inode)->nfs_client;
+
+		struct nfsd_file *nf = nfs_local_file_open(clp, hdr->cred,
+							   hdr->args.fh,
+							   hdr->args.context);
+
 		if (NFS_SERVER(hdr->inode)->nfs_client->cl_minorversion)
 			task_flags = RPC_TASK_MOVEABLE;
 		ret = nfs_initiate_pgio(NFS_CLIENT(hdr->inode),
@@ -967,7 +973,7 @@ static int nfs_generic_pg_pgios(struct nfs_pageio_descriptor *desc)
 					desc->pg_rpc_callops,
 					desc->pg_ioflags,
 					RPC_TASK_CRED_NOREF | task_flags,
-					NULL);
+					nf);
 	}
 	return ret;
 }
fs/nfs/write.c +5 -1
index 404cc5281e6a..89a49a08bc90 100644
@@ -1795,6 +1795,7 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how,
 		struct nfs_commit_info *cinfo)
 {
 	struct nfs_commit_data	*data;
+	struct nfsd_file *nf;
 	unsigned short task_flags = 0;
 
 	/* another commit raced with us */
@@ -1811,9 +1812,12 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how,
 	nfs_init_commit(data, head, NULL, cinfo);
 	if (NFS_SERVER(inode)->nfs_client->cl_minorversion)
 		task_flags = RPC_TASK_MOVEABLE;
+
+	nf = nfs_local_file_open(NFS_SERVER(inode)->nfs_client, data->cred,
+				 data->args.fh, data->context);
 	return nfs_initiate_commit(NFS_CLIENT(inode), data, NFS_PROTO(inode),
 				   data->mds_ops, how,
-				   RPC_TASK_CRED_NOREF | task_flags, NULL);
+				   RPC_TASK_CRED_NOREF | task_flags, nf);
 }
 
 /*
Signature · 1 line
2.44.0
#36 lore raw

PATCH v13 15/19 pnfs/flexfiles: enable localio support

In reply to Mike Snitzer

From: Trond Myklebust <[email protected]>

If the DS is local to this client use localio to write the data.

---
 fs/nfs/flexfilelayout/flexfilelayout.c    | 136 +++++++++++++++++++++-
 fs/nfs/flexfilelayout/flexfilelayout.h    |   2 +
 fs/nfs/flexfilelayout/flexfilelayoutdev.c |   6 +
 3 files changed, 140 insertions(+), 4 deletions(-)
fs/nfs/flexfilelayout/flexfilelayout.c +132 -4
index 01ee52551a63..d91b640f6c05 100644
@@ -11,6 +11,7 @@
 #include <linux/nfs_mount.h>
 #include <linux/nfs_page.h>
 #include <linux/module.h>
+#include <linux/file.h>
 #include <linux/sched/mm.h>
 
 #include <linux/sunrpc/metrics.h>
@@ -162,6 +163,72 @@ decode_name(struct xdr_stream *xdr, u32 *id)
 	return 0;
 }
 
+/*
+ * A dummy definition to make RCU (and non-LOCALIO compilation) happy.
+ * struct nfsd_file should never be dereferenced in this file.
+ */
+struct nfsd_file {
+       int undefined__;
+};
+
+#if IS_ENABLED(CONFIG_NFS_LOCALIO)
+
+static struct nfsd_file *
+ff_local_open_fh(struct pnfs_layout_segment *lseg,
+		 u32 ds_idx,
+		 struct nfs_client *clp,
+		 const struct cred *cred,
+		 struct nfs_fh *fh,
+		 fmode_t mode)
+{
+	struct nfs4_ff_layout_mirror *mirror = FF_LAYOUT_COMP(lseg, ds_idx);
+	struct nfsd_file *nf, *new, __rcu **pnf;
+
+	if (!nfs_server_is_local(clp))
+		return NULL;
+	if (mode & FMODE_WRITE) {
+		/*
+		 * Always request read and write access since this corresponds
+		 * to a rw layout.
+		 */
+		mode |= FMODE_READ;
+		pnf = &mirror->rw_file;
+	} else
+		pnf = &mirror->ro_file;
+
+	new = NULL;
+	rcu_read_lock();
+	nf = rcu_dereference(*pnf);
+	if (!nf) {
+		rcu_read_unlock();
+		new = nfs_local_open_fh(clp, cred, fh, mode);
+		if (IS_ERR(new))
+			return NULL;
+		rcu_read_lock();
+		/* try to swap in the pointer */
+		nf = cmpxchg(pnf, NULL, new);
+		if (!nf) {
+			nf = new;
+			new = NULL;
+		}
+	}
+	nf = nfs_to.nfsd_file_get(nf);
+	rcu_read_unlock();
+	if (new)
+		nfs_to.nfsd_file_put(new);
+	return nf;
+}
+
+#else
+static struct nfsd_file *
+ff_local_open_fh(struct pnfs_layout_segment *lseg, u32 ds_idx,
+		 struct nfs_client *clp, const struct cred *cred,
+		 struct nfs_fh *fh, fmode_t mode)
+{
+	return NULL;
+}
+#endif /* IS_ENABLED(CONFIG_NFS_LOCALIO) */
+
 static bool ff_mirror_match_fh(const struct nfs4_ff_layout_mirror *m1,
 		const struct nfs4_ff_layout_mirror *m2)
 {
@@ -237,8 +304,17 @@ static struct nfs4_ff_layout_mirror *ff_layout_alloc_mirror(gfp_t gfp_flags)
 
 static void ff_layout_free_mirror(struct nfs4_ff_layout_mirror *mirror)
 {
-	const struct cred	*cred;
+	struct nfsd_file * __maybe_unused nf;
+	const struct cred *cred;
 
+#if IS_ENABLED(CONFIG_NFS_LOCALIO)
+	nf = rcu_access_pointer(mirror->ro_file);
+	if (nf)
+		nfs_to.nfsd_file_put(nf);
+	nf = rcu_access_pointer(mirror->rw_file);
+	if (nf)
+		nfs_to.nfsd_file_put(nf);
+#endif
 	ff_layout_remove_mirror(mirror);
 	kfree(mirror->fh_versions);
 	cred = rcu_access_pointer(mirror->ro_cred);
@@ -514,6 +590,30 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh,
 		mirror = ff_layout_add_mirror(lh, fls->mirror_array[i]);
 		if (mirror != fls->mirror_array[i]) {
 			/* swap cred ptrs so free_mirror will clean up old */
+#if IS_ENABLED(CONFIG_NFS_LOCALIO)
+			if (lgr->range.iomode == IOMODE_READ) {
+				const struct cred __rcu *old =
+					xchg(&mirror->ro_cred, cred);
+				rcu_assign_pointer(fls->mirror_array[i]->ro_cred, old);
+				/* drop file if creds changed */
+				if (old != cred) {
+					struct nfsd_file *nf =
+						rcu_dereference_protected(xchg(&mirror->ro_file, NULL), 1);
+					if (nf)
+						nfs_to.nfsd_file_put(nf);
+				}
+			} else {
+				const struct cred __rcu *old =
+					xchg(&mirror->rw_cred, cred);
+				rcu_assign_pointer(fls->mirror_array[i]->rw_cred, old);
+				if (old != cred) {
+					struct nfsd_file *nf =
+						rcu_dereference_protected(xchg(&mirror->rw_file, NULL), 1);
+					if (nf)
+						nfs_to.nfsd_file_put(nf);
+				}
+			}
+#else
 			if (lgr->range.iomode == IOMODE_READ) {
 				cred = xchg(&mirror->ro_cred, cred);
 				rcu_assign_pointer(fls->mirror_array[i]->ro_cred, cred);
@@ -521,6 +621,7 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh,
 				cred = xchg(&mirror->rw_cred, cred);
 				rcu_assign_pointer(fls->mirror_array[i]->rw_cred, cred);
 			}
+#endif /* IS_ENABLED(CONFIG_NFS_LOCALIO) */
 			ff_layout_free_mirror(fls->mirror_array[i]);
 			fls->mirror_array[i] = mirror;
 		}
@@ -1756,6 +1857,7 @@ ff_layout_read_pagelist(struct nfs_pgio_header *hdr)
 	struct pnfs_layout_segment *lseg = hdr->lseg;
 	struct nfs4_pnfs_ds *ds;
 	struct rpc_clnt *ds_clnt;
+	struct nfsd_file *nf;
 	struct nfs4_ff_layout_mirror *mirror;
 	const struct cred *ds_cred;
 	loff_t offset = hdr->args.offset;
@@ -1802,11 +1904,19 @@ ff_layout_read_pagelist(struct nfs_pgio_header *hdr)
 	hdr->args.offset = offset;
 	hdr->mds_offset = offset;
 
+	/* Start IO accounting for local read */
+	nf = ff_local_open_fh(lseg, idx, ds->ds_clp, ds_cred, fh,
+			      FMODE_READ);
+	if (nf) {
+		hdr->task.tk_start = ktime_get();
+		ff_layout_read_record_layoutstats_start(&hdr->task, hdr);
+	}
+
 	/* Perform an asynchronous read to ds */
 	nfs_initiate_pgio(ds_clnt, hdr, ds_cred, ds->ds_clp->rpc_ops,
 			  vers == 3 ? &ff_layout_read_call_ops_v3 :
 				      &ff_layout_read_call_ops_v4,
-			  0, RPC_TASK_SOFTCONN, NULL);
+			  0, RPC_TASK_SOFTCONN, nf);
 	put_cred(ds_cred);
 	return PNFS_ATTEMPTED;
 
@@ -1826,6 +1936,7 @@ ff_layout_write_pagelist(struct nfs_pgio_header *hdr, int sync)
 	struct pnfs_layout_segment *lseg = hdr->lseg;
 	struct nfs4_pnfs_ds *ds;
 	struct rpc_clnt *ds_clnt;
+	struct nfsd_file *nf;
 	struct nfs4_ff_layout_mirror *mirror;
 	const struct cred *ds_cred;
 	loff_t offset = hdr->args.offset;
@@ -1870,11 +1981,19 @@ ff_layout_write_pagelist(struct nfs_pgio_header *hdr, int sync)
 	 */
 	hdr->args.offset = offset;
 
+	/* Start IO accounting for local write */
+	nf = ff_local_open_fh(lseg, idx, ds->ds_clp, ds_cred, fh,
+				FMODE_READ|FMODE_WRITE);
+	if (nf) {
+		hdr->task.tk_start = ktime_get();
+		ff_layout_write_record_layoutstats_start(&hdr->task, hdr);
+	}
+
 	/* Perform an asynchronous write */
 	nfs_initiate_pgio(ds_clnt, hdr, ds_cred, ds->ds_clp->rpc_ops,
 			  vers == 3 ? &ff_layout_write_call_ops_v3 :
 				      &ff_layout_write_call_ops_v4,
-			  sync, RPC_TASK_SOFTCONN, NULL);
+			  sync, RPC_TASK_SOFTCONN, nf);
 	put_cred(ds_cred);
 	return PNFS_ATTEMPTED;
 
@@ -1908,6 +2027,7 @@ static int ff_layout_initiate_commit(struct nfs_commit_data *data, int how)
 	struct pnfs_layout_segment *lseg = data->lseg;
 	struct nfs4_pnfs_ds *ds;
 	struct rpc_clnt *ds_clnt;
+	struct nfsd_file *nf;
 	struct nfs4_ff_layout_mirror *mirror;
 	const struct cred *ds_cred;
 	u32 idx;
@@ -1946,10 +2066,18 @@ static int ff_layout_initiate_commit(struct nfs_commit_data *data, int how)
 	if (fh)
 		data->args.fh = fh;
 
+	/* Start IO accounting for local commit */
+	nf = ff_local_open_fh(lseg, idx, ds->ds_clp, ds_cred, fh,
+			      FMODE_READ|FMODE_WRITE);
+	if (nf) {
+		data->task.tk_start = ktime_get();
+		ff_layout_commit_record_layoutstats_start(&data->task, data);
+	}
+
 	ret = nfs_initiate_commit(ds_clnt, data, ds->ds_clp->rpc_ops,
 				   vers == 3 ? &ff_layout_commit_call_ops_v3 :
 					       &ff_layout_commit_call_ops_v4,
-				   how, RPC_TASK_SOFTCONN, NULL);
+				   how, RPC_TASK_SOFTCONN, nf);
 	put_cred(ds_cred);
 	return ret;
 out_err:
fs/nfs/flexfilelayout/flexfilelayout.h +2 -0
index f84b3fb0dddd..562e7e27a8b5 100644
@@ -82,7 +82,9 @@ struct nfs4_ff_layout_mirror {
 	struct nfs_fh			*fh_versions;
 	nfs4_stateid			stateid;
 	const struct cred __rcu		*ro_cred;
+	struct nfsd_file __rcu		*ro_file;
 	const struct cred __rcu		*rw_cred;
+	struct nfsd_file __rcu		*rw_file;
 	refcount_t			ref;
 	spinlock_t			lock;
 	unsigned long			flags;
fs/nfs/flexfilelayout/flexfilelayoutdev.c +6 -0
index e028f5a0ef5f..e58bedfb1dcc 100644
@@ -395,6 +395,12 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg,
 
 	/* connect success, check rsize/wsize limit */
 	if (!status) {
+		/*
+		 * ds_clp is put in destroy_ds().
+		 * keep ds_clp even if DS is local, so that if local IO cannot
+		 * proceed somehow, we can fall back to NFS whenever we want.
+		 */
+		nfs_local_probe(ds->ds_clp);
 		max_payload =
 			nfs_block_size(rpc_max_payload(ds->ds_clp->cl_rpcclient),
 				       NULL);
Signature · 1 line
2.44.0
#37 lore raw

Re: [PATCH v13 15/19] pnfs/flexfiles: enable localio support

Quoted 25 lines · Mike Snitzer

From: Trond Myklebust <[email protected]>

If the DS is local to this client use localio to write the data.

---
 fs/nfs/flexfilelayout/flexfilelayout.c    | 136 +++++++++++++++++++++-
 fs/nfs/flexfilelayout/flexfilelayout.h    |   2 +
 fs/nfs/flexfilelayout/flexfilelayoutdev.c |   6 +
 3 files changed, 140 insertions(+), 4 deletions(-)
fs/nfs/flexfilelayout/flexfilelayout.c +1 -0
index 01ee52551a63..d91b640f6c05 100644
@@ -11,6 +11,7 @@
 #include <linux/nfs_mount.h>
 #include <linux/nfs_page.h>
 #include <linux/module.h>
+#include <linux/file.h>
 #include <linux/sched/mm.h>
 
 #include <linux/sunrpc/metrics.h>

fs/nfs/flexfilelayout/flexfilelayout.c decode_name(struct xdr_stream *xdr, u32 *id) · line 172

Quoted 10 lines
fs/nfs/flexfilelayout/flexfilelayout.c +7 -0
@@ -162,6 +163,72 @@ decode_name(struct xdr_stream *xdr, u32 *id)
 	return 0;
 }
 
+/*
+ * A dummy definition to make RCU (and non-LOCALIO compilation) happy.
+ * struct nfsd_file should never be dereferenced in this file.
+ */
+struct nfsd_file {
+       int undefined__;
+};

I removed this and tried building both with and without LOCALIO enabled and the compiler didn't complain.
Could you tell me what to do to see the unhappiness you mention?

fs/nfs/flexfilelayout/flexfilelayout.h struct nfs4_ff_layout_mirror { · line 87

Quoted 11 lines
fs/nfs/flexfilelayout/flexfilelayout.h +2 -0
index f84b3fb0dddd..562e7e27a8b5 100644
@@ -82,7 +82,9 @@ struct nfs4_ff_layout_mirror {
 	struct nfs_fh			*fh_versions;
 	nfs4_stateid			stateid;
 	const struct cred __rcu		*ro_cred;
+	struct nfsd_file __rcu		*ro_file;
 	const struct cred __rcu		*rw_cred;
+	struct nfsd_file __rcu		*rw_file;

What is the lifetime of a layout_mirror? Does it live for longer than a single IO request? If so we have a problem as this will pin the nfsd_file until the layout is returned.

Thanks,
NeilBrown

#38 lore raw

Re: [PATCH v13 15/19] pnfs/flexfiles: enable localio support

Quoted 29 lines · NeilBrown
Quoted 28 lines · Mike Snitzer

From: Trond Myklebust <[email protected]>

If the DS is local to this client use localio to write the data.

---
 fs/nfs/flexfilelayout/flexfilelayout.c    | 136 +++++++++++++++++++++-
 fs/nfs/flexfilelayout/flexfilelayout.h    |   2 +
 fs/nfs/flexfilelayout/flexfilelayoutdev.c |   6 +
 3 files changed, 140 insertions(+), 4 deletions(-)
fs/nfs/flexfilelayout/flexfilelayout.c +1 -0
index 01ee52551a63..d91b640f6c05 100644
@@ -11,6 +11,7 @@
 #include <linux/nfs_mount.h>
 #include <linux/nfs_page.h>
 #include <linux/module.h>
+#include <linux/file.h>
 #include <linux/sched/mm.h>
 
 #include <linux/sunrpc/metrics.h>
@@ -162,6 +163,72 @@ decode_name(struct xdr_stream *xdr, u32 *id)
 	return 0;
 }
 
Quoted 11 lines

fs/nfs/flexfilelayout/flexfilelayout.c decode_name(struct xdr_stream *xdr, u32 *id) · line 172

Quoted 7 lines
fs/nfs/flexfilelayout/flexfilelayout.c +7 -0
+/*
+ * A dummy definition to make RCU (and non-LOCALIO compilation) happy.
+ * struct nfsd_file should never be dereferenced in this file.
+ */
+struct nfsd_file {
+       int undefined__;
+};

I removed this and tried building both with and without LOCALIO enabled and the compiler didn't complain.
Could you tell me what to do to see the unhappiness you mention?

Sorry, I can remove the dummy definition for upstream. That was leftover from the backport I did to 5.15.y stable@ kernel. Older kernels' RCU code dereferences what should just be an opaque pointer and (ab)use typeof. So without the dummy definition compiling against 5.15.y fails with:

CC [M] fs/nfs/flexfilelayout/flexfilelayout.o
In file included from ./include/linux/rbtree.h:24,

                 from ./include/linux/mm_types.h:10,
                 from ./include/linux/mmzone.h:21,
                 from ./include/linux/gfp.h:6,
                 from ./include/linux/mm.h:10,
                 from ./include/linux/nfs_fs.h:23,
                 from fs/nfs/flexfilelayout/flexfilelayout.c:10:
fs/nfs/flexfilelayout/flexfilelayout.c: In function `ff_local_open_fh´:
./include/linux/rcupdate.h:441:9: error: dereferencing pointer to incomplete type `struct nfsd_file´
  typeof(*p) *________p1 = (typeof(*p) *__force)READ_ONCE(p); \
         ^

./include/linux/rcupdate.h:580:2: note: in expansion of macro `__rcu_dereference_check´
__rcu_dereference_check((p), (c) || rcu_read_lock_held(), __rcu) ^~~~~~~~~~~~~~~~~~~~~~~
./include/linux/rcupdate.h:648:28: note: in expansion of macro `rcu_dereference_check´

 #define rcu_dereference(p) rcu_dereference_check(p, 0)
                            ^~~~~~~~~~~~~~~~~~~~~
fs/nfs/flexfilelayout/flexfilelayout.c:193:7: note: in expansion of macro `rcu_dereference´
  nf = rcu_dereference(*pnf);
       ^~~~~~~~~~~~~~~
Quoted 7 lines
Quoted 7 lines
fs/nfs/flexfilelayout/flexfilelayout.h +0 -0
index f84b3fb0dddd..562e7e27a8b5 100644
@@ -82,7 +82,9 @@ struct nfs4_ff_layout_mirror {
 	struct nfs_fh			*fh_versions;
 	nfs4_stateid			stateid;
Quoted 8 lines

fs/nfs/flexfilelayout/flexfilelayout.h struct nfs4_ff_layout_mirror { · line 87

Quoted 4 lines
fs/nfs/flexfilelayout/flexfilelayout.h +2 -0
 	const struct cred __rcu		*ro_cred;
+	struct nfsd_file __rcu		*ro_file;
 	const struct cred __rcu		*rw_cred;
+	struct nfsd_file __rcu		*rw_file;

What is the lifetime of a layout_mirror? Does it live for longer than a single IO request? If so we have a problem as this will pin the nfsd_file until the layout is returned.

Ah, yeah lifetime is longer than an IO... so we have the issue of pnfs (flexfileslayout) holding nfsd_files open in the client; which will prevent backing filesystem from being unmounted. I haven't done that same unmount test (which you reported I fixed for normal NFS) against pNFS with flexfiles. Will sort it out.

#39 lore raw

Re: [PATCH v13 15/19] pnfs/flexfiles: enable localio support

Quoted 64 lines · Mike Snitzer
Quoted 29 lines · NeilBrown
Quoted 28 lines · Mike Snitzer

From: Trond Myklebust <[email protected]>

If the DS is local to this client use localio to write the data.

---
 fs/nfs/flexfilelayout/flexfilelayout.c    | 136 +++++++++++++++++++++-
 fs/nfs/flexfilelayout/flexfilelayout.h    |   2 +
 fs/nfs/flexfilelayout/flexfilelayoutdev.c |   6 +
 3 files changed, 140 insertions(+), 4 deletions(-)
fs/nfs/flexfilelayout/flexfilelayout.c +1 -0
index 01ee52551a63..d91b640f6c05 100644
@@ -11,6 +11,7 @@
 #include <linux/nfs_mount.h>
 #include <linux/nfs_page.h>
 #include <linux/module.h>
+#include <linux/file.h>
 #include <linux/sched/mm.h>
 
 #include <linux/sunrpc/metrics.h>
@@ -162,6 +163,72 @@ decode_name(struct xdr_stream *xdr, u32 *id)
 	return 0;
 }
 
Quoted 11 lines

fs/nfs/flexfilelayout/flexfilelayout.c decode_name(struct xdr_stream *xdr, u32 *id) · line 172

Quoted 7 lines
fs/nfs/flexfilelayout/flexfilelayout.c +7 -0
+/*
+ * A dummy definition to make RCU (and non-LOCALIO compilation) happy.
+ * struct nfsd_file should never be dereferenced in this file.
+ */
+struct nfsd_file {
+       int undefined__;
+};

I removed this and tried building both with and without LOCALIO enabled and the compiler didn't complain.
Could you tell me what to do to see the unhappiness you mention?

Sorry, I can remove the dummy definition for upstream. That was leftover from the backport I did to 5.15.y stable@ kernel. Older kernels' RCU code dereferences what should just be an opaque pointer and (ab)use typeof. So without the dummy definition compiling against 5.15.y fails with:

CC [M] fs/nfs/flexfilelayout/flexfilelayout.o
In file included from ./include/linux/rbtree.h:24,

                 from ./include/linux/mm_types.h:10,
                 from ./include/linux/mmzone.h:21,
                 from ./include/linux/gfp.h:6,
                 from ./include/linux/mm.h:10,
                 from ./include/linux/nfs_fs.h:23,
                 from fs/nfs/flexfilelayout/flexfilelayout.c:10:
fs/nfs/flexfilelayout/flexfilelayout.c: In function `ff_local_open_fh´:
./include/linux/rcupdate.h:441:9: error: dereferencing pointer to incomplete type `struct nfsd_file´
  typeof(*p) *________p1 = (typeof(*p) *__force)READ_ONCE(p); \
         ^

./include/linux/rcupdate.h:580:2: note: in expansion of macro `__rcu_dereference_check´
__rcu_dereference_check((p), (c) || rcu_read_lock_held(), __rcu) ^~~~~~~~~~~~~~~~~~~~~~~
./include/linux/rcupdate.h:648:28: note: in expansion of macro `rcu_dereference_check´

Quoted 5 lines
 #define rcu_dereference(p) rcu_dereference_check(p, 0)
                            ^~~~~~~~~~~~~~~~~~~~~
fs/nfs/flexfilelayout/flexfilelayout.c:193:7: note: in expansion of macro `rcu_dereference´
  nf = rcu_dereference(*pnf);
       ^~~~~~~~~~~~~~~

Ahhh - good to know. Thanks!

NeilBrown

#40 lore raw

PATCH v13 16/19 nfs/localio: use dedicated workqueues for filesystem read and write

In reply to Mike Snitzer

From: Trond Myklebust <[email protected]>

For localio access, don't call filesystem read() and write() routines directly. This solves two problems:

1) localio writes need to use a normal (non-memreclaim) unbound workqueue. This avoids imposing new requirements on how underlying filesystems process frontend IO, which would cause a large amount of work to update all filesystems. Without this change, when XFS starts getting low on space, XFS flushes work on a non-memreclaim work queue, which causes a priority inversion problem:

00573 workqueue: WQ_MEM_RECLAIM writeback:wb_workfn is flushing !WQ_MEM_RECLAIM xfs-sync/vdc:xfs_flush_inodes_worker 00573 WARNING: CPU: 6 PID: 8525 at kernel/workqueue.c:3706 check_flush_dependency+0x2a4/0x328 00573 Modules linked in:
00573 CPU: 6 PID: 8525 Comm: kworker/u71:5 Not tainted 6.10.0-rc3-ktest-00032-g2b0a133403ab #18502 00573 Hardware name: linux,dummy-virt (DT)
00573 Workqueue: writeback wb_workfn (flush-0:33)
00573 pstate: 400010c5 (nZcv daIF -PAN -UAO -TCO -DIT +SSBS BTYPE=--)
00573 pc : check_flush_dependency+0x2a4/0x328
00573 lr : check_flush_dependency+0x2a4/0x328
00573 sp : ffff0000c5f06bb0
00573 x29: ffff0000c5f06bb0 x28: ffff0000c998a908 x27: 1fffe00019331521
00573 x26: ffff0000d0620900 x25: ffff0000c5f06ca0 x24: ffff8000828848c0
00573 x23: 1fffe00018be0d8e x22: ffff0000c1210000 x21: ffff0000c75fde00
00573 x20: ffff800080bfd258 x19: ffff0000cad63400 x18: ffff0000cd3a4810
00573 x17: 0000000000000000 x16: 0000000000000000 x15: ffff800080508d98
00573 x14: 0000000000000000 x13: 204d49414c434552 x12: 1fffe0001b6eeab2
00573 x11: ffff60001b6eeab2 x10: dfff800000000000 x9 : ffff60001b6eeab3
00573 x8 : 0000000000000001 x7 : 00009fffe491154e x6 : ffff0000db775593
00573 x5 : ffff0000db775590 x4 : ffff0000db775590 x3 : 0000000000000000
00573 x2 : 0000000000000027 x1 : ffff600018be0d62 x0 : dfff800000000000
00573 Call trace:

00573  check_flush_dependency+0x2a4/0x328
00573  __flush_work+0x184/0x5c8
00573  flush_work+0x18/0x28
00573  xfs_flush_inodes+0x68/0x88
00573  xfs_file_buffered_write+0x128/0x6f0
00573  xfs_file_write_iter+0x358/0x448
00573  nfs_local_doio+0x854/0x1568
00573  nfs_initiate_pgio+0x214/0x418
00573  nfs_generic_pg_pgios+0x304/0x480
00573  nfs_pageio_doio+0xe8/0x240
00573  nfs_pageio_complete+0x160/0x480
00573  nfs_writepages+0x300/0x4f0
00573  do_writepages+0x12c/0x4a0
00573  __writeback_single_inode+0xd4/0xa68
00573  writeback_sb_inodes+0x470/0xcb0
00573  __writeback_inodes_wb+0xb0/0x1d0
00573  wb_writeback+0x594/0x808
00573  wb_workfn+0x5e8/0x9e0
00573  process_scheduled_works+0x53c/0xd90
00573  worker_thread+0x370/0x8c8
00573  kthread+0x258/0x2e8
00573  ret_from_fork+0x10/0x20

2) Some filesystem writeback routines can end up taking up a lot of stack space (particularly XFS). Instead of risking running over due to the extra overhead from the NFS stack, we should just call these routines from a workqueue job. Since we need to do this to address 1) above we're able to avoid possibly blowing the stack "for free".

Use of dedicated workqueues improves performance over using the system_unbound_wq.

Also, the creds used to open the file are used to override_creds() in both nfs_local_call_read() and nfs_local_call_write() -- otherwise the workqueue could have elevated capabilities (which the caller may not).

Lastly, care is taken to set PF_LOCAL_THROTTLE | PF_MEMALLOC_NOIO in nfs_do_local_write() to avoid writeback deadlocks.

The PF_LOCAL_THROTTLE flag prevents deadlocks in balance_dirty_pages() by causing writes to only be throttled against other writes to the same bdi (it keeps the throttling local). Normally all writes to bdi(s) are throttled equally (after throughput factors are allowed for).

The PF_MEMALLOC_NOIO flag prevents the lower filesystem IO from causing memory reclaim to re-enter filesystems or IO devices and so prevents deadlocks from occuring where IO that cleans pages is waiting on IO to complete.

---
 fs/nfs/inode.c    | 57 ++++++++++++++++++++++-----------
 fs/nfs/internal.h |  1 +
 fs/nfs/localio.c  | 81 ++++++++++++++++++++++++++++++++---------------
 3 files changed, 95 insertions(+), 44 deletions(-)
fs/nfs/inode.c +38 -19
index b4914a11c3c2..542c7d97b235 100644
@@ -2461,35 +2461,54 @@ static void nfs_destroy_inodecache(void)
 	kmem_cache_destroy(nfs_inode_cachep);
 }
 
+struct workqueue_struct *nfslocaliod_workqueue;
 struct workqueue_struct *nfsiod_workqueue;
 EXPORT_SYMBOL_GPL(nfsiod_workqueue);
 
 /*
- * start up the nfsiod workqueue
- */
-static int nfsiod_start(void)
-{
-	struct workqueue_struct *wq;
-	dprintk("RPC:       creating workqueue nfsiod\n");
-	wq = alloc_workqueue("nfsiod", WQ_MEM_RECLAIM | WQ_UNBOUND, 0);
-	if (wq == NULL)
-		return -ENOMEM;
-	nfsiod_workqueue = wq;
-	return 0;
-}
-
-/*
- * Destroy the nfsiod workqueue
+ * Destroy the nfsiod workqueues
  */
 static void nfsiod_stop(void)
 {
 	struct workqueue_struct *wq;
 
 	wq = nfsiod_workqueue;
-	if (wq == NULL)
-		return;
-	nfsiod_workqueue = NULL;
-	destroy_workqueue(wq);
+	if (wq != NULL) {
+		nfsiod_workqueue = NULL;
+		destroy_workqueue(wq);
+	}
+#if IS_ENABLED(CONFIG_NFS_LOCALIO)
+	wq = nfslocaliod_workqueue;
+	if (wq != NULL) {
+		nfslocaliod_workqueue = NULL;
+		destroy_workqueue(wq);
+	}
+#endif /* CONFIG_NFS_LOCALIO */
+}
+
+/*
+ * Start the nfsiod workqueues
+ */
+static int nfsiod_start(void)
+{
+	dprintk("RPC:       creating workqueue nfsiod\n");
+	nfsiod_workqueue = alloc_workqueue("nfsiod", WQ_MEM_RECLAIM | WQ_UNBOUND, 0);
+	if (nfsiod_workqueue == NULL)
+		return -ENOMEM;
+#if IS_ENABLED(CONFIG_NFS_LOCALIO)
+	/*
+	 * localio writes need to use a normal (non-memreclaim) workqueue.
+	 * When we start getting low on space, XFS goes and calls flush_work() on
+	 * a non-memreclaim work queue, which causes a priority inversion problem.
+	 */
+	dprintk("RPC:       creating workqueue nfslocaliod\n");
+	nfslocaliod_workqueue = alloc_workqueue("nfslocaliod", WQ_UNBOUND, 0);
+	if (unlikely(nfslocaliod_workqueue == NULL)) {
+		nfsiod_stop();
+		return -ENOMEM;
+	}
+#endif /* CONFIG_NFS_LOCALIO */
+	return 0;
 }
 
 unsigned int nfs_net_id;
fs/nfs/internal.h +1 -0
index 0f603723504b..a7677a16e929 100644
@@ -439,6 +439,7 @@ int nfs_check_flags(int);
 
 /* inode.c */
 extern struct workqueue_struct *nfsiod_workqueue;
+extern struct workqueue_struct *nfslocaliod_workqueue;
 extern struct inode *nfs_alloc_inode(struct super_block *sb);
 extern void nfs_free_inode(struct inode *);
 extern int nfs_write_inode(struct inode *, struct writeback_control *);
fs/nfs/localio.c +56 -25

123 lines not on this page · raw message

Signature · 1 line
2.44.0
#41 lore raw

PATCH v13 17/19 nfs: implement client support for NFS_LOCALIO_PROGRAM

In reply to Mike Snitzer

The LOCALIO auxiliary RPC protocol consists of a single "UUID_IS_LOCAL" RPC method that allows the Linux NFS client to verify the local Linux NFS server can see the nonce (single-use UUID) the client generated and made available in nfs_common for subsequent lookup and verification by the NFS server. If matched, the NFS server populates members in the nfs_uuid_t struct. The NFS client then transfers these nfs_uuid_t struct member pointers to the nfs_client struct and cleans up the nfs_uuid_t struct. See: fs/nfs/localio.c:nfs_local_probe()

This protocol isn't part of an IETF standard, nor does it need to be considering it is Linux-to-Linux auxiliary RPC protocol that amounts to an implementation detail.

Localio is only supported when UNIX-style authentication (AUTH_UNIX, aka AUTH_SYS) is used (enforced by fs/nfs/localio.c:nfs_local_probe()).

The UUID_IS_LOCAL method encodes the client generated uuid_t in terms of the fixed UUID_SIZE (16 bytes). The fixed size opaque encode and decode XDR methods are used instead of the less efficient variable sized methods.

Having a nonce (single-use uuid) is better than using the same uuid for the life of the server, and sending it proactively by client rather than reactively by the server is also safer.

[NeilBrown factored out and simplified a single localio protocol and proposed making the uuid short-lived]

---
 fs/nfs/client.c  |   6 +-
 fs/nfs/localio.c | 145 +++++++++++++++++++++++++++++++++++++++++++++--
 2 files changed, 144 insertions(+), 7 deletions(-)
fs/nfs/client.c +4 -2
index 95e8d76bd49f..fe60a82f06d8 100644
@@ -434,8 +434,10 @@ struct nfs_client *nfs_get_client(const struct nfs_client_initdata *cl_init)
 			list_add_tail(&new->cl_share_link,
 					&nn->nfs_client_list);
 			spin_unlock(&nn->nfs_client_lock);
-			nfs_local_probe(new);
-			return rpc_ops->init_client(new, cl_init);
+			new = rpc_ops->init_client(new, cl_init);
+			if (!IS_ERR(new))
+				 nfs_local_probe(new);
+			return new;
 		}
 
 		spin_unlock(&nn->nfs_client_lock);
fs/nfs/localio.c +140 -5

182 lines not on this page · raw message

Signature · 1 line
2.44.0
#42 lore raw

PATCH v13 18/19 nfs: add Documentation/filesystems/nfs/localio.rst

In reply to Mike Snitzer

This document gives an overview of the LOCALIO auxiliary RPC protocol added to the Linux NFS client and server to allow them to reliably handshake to determine if they are on the same host.

Once an NFS client and server handshake as "local", the client will bypass the network RPC protocol for read, write and commit operations. Due to this XDR and RPC bypass, these operations will operate faster.

---
 Documentation/filesystems/nfs/localio.rst | 199 ++++++++++++++++++++++
 1 file changed, 199 insertions(+)
 create mode 100644 Documentation/filesystems/nfs/localio.rst
Documentation/filesystems/nfs/localio.rst new file +199 -0
new file mode 100644
index 000000000000..8cceb3db386a
@@ -0,0 +1,199 @@
+===========
+NFS LOCALIO
+===========
+
+Overview
+========
+
+The LOCALIO auxiliary RPC protocol allows the Linux NFS client and
+server to reliably handshake to determine if they are on the same host.
+
+Once an NFS client and server handshake as "local", the client will
+bypass the network RPC protocol for read, write and commit operations.
+Due to this XDR and RPC bypass, these operations will operate faster.
+
+The LOCALIO auxiliary protocol's implementation, which uses the same
+connection as NFS traffic, follows the pattern established by the NFS
+ACL protocol extension.
+
+The LOCALIO auxiliary protocol is needed to allow robust discovery of
+clients local to their servers. In a private implementation that
+preceded use of this LOCALIO protocol, a fragile sockaddr network
+address based match against all local network interfaces was attempted.
+But unlike the LOCALIO protocol, the sockaddr-based matching didn't
+handle use of iptables or containers.
+
+The robust handshake between local client and server is just the
+beginning, the ultimate use case this locality makes possible is the
+client is able to open files and issue reads, writes and commits
+directly to the server without having to go over the network. The
+requirement is to perform these loopback NFS operations as efficiently
+as possible, this is particularly useful for container use cases
+(e.g. kubernetes) where it is possible to run an IO job local to the
+server.
+
+The performance advantage realized from LOCALIO's ability to bypass
+using XDR and RPC for reads, writes and commits can be extreme, e.g.:
+
+fio for 20 secs with directio, qd of 8, 16 libaio threads:
+- With LOCALIO:
+  4K read:    IOPS=979k,  BW=3825MiB/s (4011MB/s)(74.7GiB/20002msec)
+  4K write:   IOPS=165k,  BW=646MiB/s  (678MB/s)(12.6GiB/20002msec)
+  128K read:  IOPS=402k,  BW=49.1GiB/s (52.7GB/s)(982GiB/20002msec)
+  128K write: IOPS=11.5k, BW=1433MiB/s (1503MB/s)(28.0GiB/20004msec)
+
+- Without LOCALIO:
+  4K read:    IOPS=79.2k, BW=309MiB/s  (324MB/s)(6188MiB/20003msec)
+  4K write:   IOPS=59.8k, BW=234MiB/s  (245MB/s)(4671MiB/20002msec)
+  128K read:  IOPS=33.9k, BW=4234MiB/s (4440MB/s)(82.7GiB/20004msec)
+  128K write: IOPS=11.5k, BW=1434MiB/s (1504MB/s)(28.0GiB/20011msec)
+
+fio for 20 secs with directio, qd of 8, 1 libaio thread:
+- With LOCALIO:
+  4K read:    IOPS=230k,  BW=898MiB/s  (941MB/s)(17.5GiB/20001msec)
+  4K write:   IOPS=22.6k, BW=88.3MiB/s (92.6MB/s)(1766MiB/20001msec)
+  128K read:  IOPS=38.8k, BW=4855MiB/s (5091MB/s)(94.8GiB/20001msec)
+  128K write: IOPS=11.4k, BW=1428MiB/s (1497MB/s)(27.9GiB/20001msec)
+
+- Without LOCALIO:
+  4K read:    IOPS=77.1k, BW=301MiB/s  (316MB/s)(6022MiB/20001msec)
+  4K write:   IOPS=32.8k, BW=128MiB/s  (135MB/s)(2566MiB/20001msec)
+  128K read:  IOPS=24.4k, BW=3050MiB/s (3198MB/s)(59.6GiB/20001msec)
+  128K write: IOPS=11.4k, BW=1430MiB/s (1500MB/s)(27.9GiB/20001msec)
+
+RPC
+===
+
+The LOCALIO auxiliary RPC protocol consists of a single "UUID_IS_LOCAL"
+RPC method that allows the Linux NFS client to verify the local Linux
+NFS server can see the nonce (single-use UUID) the client generated and
+made available in nfs_common. This protocol isn't part of an IETF
+standard, nor does it need to be considering it is Linux-to-Linux
+auxiliary RPC protocol that amounts to an implementation detail.
+
+The UUID_IS_LOCAL method encodes the client generated uuid_t in terms of
+the fixed UUID_SIZE (16 bytes). The fixed size opaque encode and decode
+XDR methods are used instead of the less efficient variable sized
+methods.
+
+The RPC program number for the NFS_LOCALIO_PROGRAM is 400122 (as assigned
+by IANA, see https://www.iana.org/assignments/rpc-program-numbers/ ):
+Linux Kernel Organization       400122  nfslocalio
+
+The LOCALIO protocol spec in rpcgen syntax is:
+
+/* raw RFC 9562 UUID */
+#define UUID_SIZE 16
+typedef u8 uuid_t<UUID_SIZE>;
+
+program NFS_LOCALIO_PROGRAM {
+    version LOCALIO_V1 {
+        void
+            NULL(void) = 0;
+
+        void
+            UUID_IS_LOCAL(uuid_t) = 1;
+    } = 1;
+} = 400122;
+
+LOCALIO uses the same transport connection as NFS traffic. As such,
+LOCALIO is not registered with rpcbind.
+
+NFS Common and Client/Server Handshake
+======================================
+
+fs/nfs_common/nfslocalio.c provides interfaces that enable an NFS client
+to generate a nonce (single-use UUID) and associated short-lived
+nfs_uuid_t struct, register it with nfs_common for subsequent lookup and
+verification by the NFS server and if matched the NFS server populates
+members in the nfs_uuid_t struct. The nfs client then transfers these
+nfs_uuid_t struct member pointers to the nfs_client struct and cleans up
+the nfs_uuid_t struct.  See: fs/nfs/localio.c:nfs_local_probe()
+
+nfs_common's nfs_uuids list is the basis for LOCALIO enablement, as such
+it has members that point to nfsd memory for direct use by the client
+(e.g. 'net' is the server's network namespace, through it the client can
+access nn->nfsd_serv with proper rcu read access). It is this client
+and server synchronization that enables advanced usage and lifetime of
+objects to span from the host kernel's nfsd to per-container knfsd
+instances that are connected to nfs client's running on the same local
+host.
+
+NFS Client issues IO instead of Server
+======================================
+
+Because LOCALIO is focused on protocol bypass to achieve improved IO
+performance alternatives to traditional NFS wire protocol (SUNRPC with
+XDR) to access the backing filesystem must be provided.
+
+See fs/nfs/localio.c:nfs_local_open_fh() and
+fs/nfsd/localio.c:nfsd_open_local_fh() for the interface that makes
+focused use of select nfs server objects to allow a client local to a
+server to open a file pointer without needing to go over the network.
+
+The client's fs/nfs/localio.c:nfs_local_open_fh() will call into the
+server's fs/nfsd/localio.c:nfsd_open_local_fh() and carefully access
+both the nfsd network namespace and the associated nn->nfsd_serv in
+terms of RCU. If nfsd_open_local_fh() finds that client no longer sees
+valid nfsd objects (be it struct net or nn->nfsd_serv) it returns ENXIO
+to nfs_local_open_fh() and the client will try to reestablish the
+LOCALIO resources needed by calling nfs_local_probe() again. This
+recovery is needed if/when an nfsd instance running in a container were
+to reboot while a LOCALIO client is connected to it.
+
+Once the client has an open file pointer it will issue reads, writes and
+commits directly to the underlying local filesystem (normally done by
+the nfs server). As such, for these operations, the NFS client is
+issuing IO to the underlying local filesystem that it is sharing with
+the NFS server. See: fs/nfs/localio.c:nfs_local_doio() and
+fs/nfs/localio.c:nfs_local_commit().
+
+Security
+========
+
+Localio is only supported when UNIX-style authentication (AUTH_UNIX, aka
+AUTH_SYS) is used.
+
+Care is taken to ensure the same NFS security mechanisms are used
+(authentication, etc) regardless of whether LOCALIO or regular NFS
+access is used. The auth_domain established as part of the traditional
+NFS client access to the NFS server is also used for LOCALIO.
+
+Relative to containers, LOCALIO gives the client access to the network
+namespace the server has. This is required to allow the client to access
+the server's per-namespace nfsd_net struct. With traditional NFS, the
+client is afforded this same level of access (albeit in terms of the NFS
+protocol via SUNRPC). No other namespaces (user, mount, etc) have been
+altered or purposely extended from the server to the client.
+
+Testing
+=======
+
+The LOCALIO auxiliary protocol and associated NFS LOCALIO read, write
+and commit access have proven stable against various test scenarios:
+
+- Client and server both on the same host.
+
+- All permutations of client and server support enablement for both
+  local and remote client and server.
+
+- Testing against NFS storage products that don't support the LOCALIO
+  protocol was also performed.
+
+- Client on host, server within a container (for both v3 and v4.2).
+  The container testing was in terms of podman managed containers and
+  includes successful container stop/restart scenario.
+
+- Formalizing these test scenarios in terms of existing test
+  infrastructure is on-going. Initial regular coverage is provided in
+  terms of ktest running xfstests against a LOCALIO-enabled NFS loopback
+  mount configuration, and includes lockdep and KASAN coverage, see:
+  https://evilpiepirate.org/~testdashboard/ci?user=snitzer&branch=snitm-nfs-next
+  https://github.com/koverstreet/ktest
+
+- Various kdevops testing (in terms of "Chuck's BuildBot") has been
+  performed to regularly verify the LOCALIO changes haven't caused any
+  regressions to non-LOCALIO NFS use cases.
+
+- All of Hammerspace's various sanity tests pass with LOCALIO enabled
+  (this includes numerous pNFS and flexfiles tests).
Signature · 1 line
2.44.0
#43 lore raw

PATCH v13 19/19 nfs: add FAQ section to Documentation/filesystems/nfs/localio.rst

In reply to Mike Snitzer

From: Trond Myklebust <[email protected]>

Add a FAQ section to give answers to questions that have been raised during review of the localio feature.

---
 Documentation/filesystems/nfs/localio.rst | 77 +++++++++++++++++++++++
 1 file changed, 77 insertions(+)
Documentation/filesystems/nfs/localio.rst +77 -0
index 8cceb3db386a..4b6d63246479 100644
@@ -61,6 +61,83 @@ fio for 20 secs with directio, qd of 8, 1 libaio thread:
   128K read:  IOPS=24.4k, BW=3050MiB/s (3198MB/s)(59.6GiB/20001msec)
   128K write: IOPS=11.4k, BW=1430MiB/s (1500MB/s)(27.9GiB/20001msec)
 
+FAQ
+===
+
+1. What are the use cases for LOCALIO?
+
+   a. Workloads where the NFS client and server are on the same host
+      realize improved IO performance. In particular, it is common when
+      running containerised workloads for jobs to find themselves
+      running on the same host as the knfsd server being used for
+      storage.
+
+2. What are the requirements for LOCALIO?
+
+   a. Bypass use of the network RPC protocol as much as possible. This
+      includes bypassing XDR and RPC for open, read, write and commit
+      operations.
+   b. Allow client and server to autonomously discover if they are
+      running local to each other without making any assumptions about
+      the local network topology.
+   c. Support the use of containers by being compatible with relevant
+      namespaces (e.g. network, user, mount).
+   d. Support all versions of NFS. NFSv3 is of particular importance
+      because it has wide enterprise usage and pNFS flexfiles makes use
+      of it for the data path.
+
+3. Why doesn’t LOCALIO just compare IP addresses or hostnames when
+   deciding if the NFS client and server are co-located on the same
+   host?
+
+   Since one of the main use cases is containerised workloads, we cannot
+   assume that IP addresses will be shared between the client and
+   server. This sets up a requirement for a handshake protocol that
+   needs to go over the same connection as the NFS traffic in order to
+   identify that the client and the server really are running on the
+   same host. The handshake uses a secret that is sent over the wire,
+   and can be verified by both parties by comparing with a value stored
+   in shared kernel memory if they are truly co-located.
+
+4. Does LOCALIO improve pNFS flexfiles?
+
+   Yes, LOCALIO complements pNFS flexfiles by allowing it to take
+   advantage of NFS client and server locality.  Policy that initiates
+   client IO as closely to the server where the data is stored naturally
+   benefits from the data path optimization LOCALIO provides.
+
+5. Why not develop a new pNFS layout to enable LOCALIO?
+
+   A new pNFS layout could be developed, but doing so would put the
+   onus on the server to somehow discover that the client is co-located
+   when deciding to hand out the layout.
+   There is value in a simpler approach (as provided by LOCALIO) that
+   allows the NFS client to negotiate and leverage locality without
+   requiring more elaborate modeling and discovery of such locality in a
+   more centralized manner.
+
+6. Why is having the client perform a server-side file OPEN, without
+   using RPC, beneficial?  Is the benefit pNFS specific?
+
+   Avoiding the use of XDR and RPC for file opens is beneficial to
+   performance regardless of whether pNFS is used. However adding a
+   requirement to go over the wire to do an open and/or close ends up
+   negating any benefit of avoiding the wire for doing the I/O itself
+   when we’re dealing with small files. There is no benefit to replacing
+   the READ or WRITE with a new open and/or close operation that still
+   needs to go over the wire.
+
+7. Why is LOCALIO only supported with UNIX Authentication (AUTH_UNIX)?
+
+   Strong authentication is usually tied to the connection itself. It
+   works by establishing a context that is cached by the server, and
+   that acts as the key for discovering the authorisation token, which
+   can then be passed to rpc.mountd to complete the authentication
+   process. On the other hand, in the case of AUTH_UNIX, the credential
+   that was passed over the wire is used directly as the key in the
+   upcall to rpc.mountd. This simplifies the authentication process, and
+   so makes AUTH_UNIX easier to support.
+
 RPC
 ===
 
Signature · 1 line
2.44.0
#44 lore raw

Re: [PATCH v13 19/19] nfs: add FAQ section to Documentation/filesystems/nfs/localio.rst

Documentation/filesystems/nfs/localio.rst fio for 20 secs with directio, qd of 8, 1 libaio thread: · line 128

Quoted 11 lines · Mike Snitzer
Documentation/filesystems/nfs/localio.rst +11 -0
+
+6. Why is having the client perform a server-side file OPEN, without
+   using RPC, beneficial?  Is the benefit pNFS specific?
+
+   Avoiding the use of XDR and RPC for file opens is beneficial to
+   performance regardless of whether pNFS is used. However adding a
+   requirement to go over the wire to do an open and/or close ends up
+   negating any benefit of avoiding the wire for doing the I/O itself
+   when we’re dealing with small files. There is no benefit to replacing
+   the READ or WRITE with a new open and/or close operation that still
+   needs to go over the wire.

I don't think the above is correct.
The current code still does a normal NFSv4 OPEN or NFSv3 GETATTR when then client opens a file. Only the READ/WRITE/COMMIT operations are avoided.

While I'm not advocating for an over-the-wire request to map a filehandle to a struct nfsd_file*, I don't think you can convincingly argue against it without concrete performance measurements.

NeilBrown

#45 lore raw

Re: [PATCH v13 19/19] nfs: add FAQ section to Documentation/filesystems/nfs/localio.rst

Quoted 18 lines · NeilBrown

Documentation/filesystems/nfs/localio.rst fio for 20 secs with directio, qd of 8, 1 libaio thread: · line 128

Quoted 11 lines · Mike Snitzer
Documentation/filesystems/nfs/localio.rst +11 -0
+
+6. Why is having the client perform a server-side file OPEN, without
+   using RPC, beneficial?  Is the benefit pNFS specific?
+
+   Avoiding the use of XDR and RPC for file opens is beneficial to
+   performance regardless of whether pNFS is used. However adding a
+   requirement to go over the wire to do an open and/or close ends up
+   negating any benefit of avoiding the wire for doing the I/O itself
+   when we’re dealing with small files. There is no benefit to replacing
+   the READ or WRITE with a new open and/or close operation that still
+   needs to go over the wire.

I don't think the above is correct.

I struggled with this text too.

I thought the reason we want a server-side file OPEN is so that proper access authorization, same as would be done on a remote access, can be done.

Quoted 9 lines

The current code still does a normal NFSv4 OPEN or NFSv3 GETATTR when then client opens a file. Only the READ/WRITE/COMMIT operations are avoided.

While I'm not advocating for an over-the-wire request to map a filehandle to a struct nfsd_file*, I don't think you can convincingly argue against it without concrete performance measurements.

NeilBrown

--
Chuck Lever

#46 lore raw

Re: [PATCH v13 19/19] nfs: add FAQ section to Documentation/filesystems/nfs/localio.rst

Quoted 26 lines · Chuck Lever III
Quoted 17 lines · NeilBrown
Quoted 12 lines · Mike Snitzer

+
+6. Why is having the client perform a server-side file OPEN, without
+   using RPC, beneficial?  Is the benefit pNFS specific? +
+   Avoiding the use of XDR and RPC for file opens is beneficial to
+   performance regardless of whether pNFS is used. However adding a
+   requirement to go over the wire to do an open and/or close ends up
+   negating any benefit of avoiding the wire for doing the I/O

Quoted 8 lines
Quoted 6 lines

itself
+   when we’re dealing with small files. There is no benefit to replacing
+   the READ or WRITE with a new open and/or close operation that still
+   needs to go over the wire.

I don't think the above is correct.

Quoted 6 lines

I struggled with this text too.

I thought the reason we want a server-side file OPEN is so that proper access authorization, same as would be done on a remote access, can be done.

You're conflating "server-side file open" with "on the wire open". The code does do a server side file open, and does call up to rpc.mountd to authenticate the client's IP address and domain.

The text is basically pointing out that if you have to add stateful on- the-wire operations for small files (e.g. size < 1MB), then you might as well just send the READ or WRITE instead.

Quoted 7 lines
Quoted 7 lines

The current code still does a normal NFSv4 OPEN or NFSv3 GETATTR when
then client opens a file.  Only the READ/WRITE/COMMIT operations are
avoided.

Quoted 5 lines
Quoted 4 lines

While I'm not advocating for an over-the-wire request to map a filehandle to a struct nfsd_file*, I don't think you can convincingly
argue against it without concrete performance measurements.

What is the value of doing an open over the wire? What are you trying to accomplish that can't be accomplished without going over the wire?

Signature · 3 lines
Trond Myklebust
Linux NFS client maintainer, Hammerspace
[email protected]
#47 lore raw

Re: [PATCH v13 19/19] nfs: add FAQ section to Documentation/filesystems/nfs/localio.rst

Quoted 4 lines · Trond Myklebust
Quoted 4 lines
Quoted 4 lines · NeilBrown
Quoted 8 lines
Quoted 5 lines
Quoted 4 lines

While I'm not advocating for an over-the-wire request to map a filehandle to a struct nfsd_file*, I don't think you can convincingly
argue against it without concrete performance measurements.

What is the value of doing an open over the wire? What are you trying to accomplish that can't be accomplished without going over the wire?

The advantage of going over the wire is avoiding code duplication. The cost is latency. Obviously the goal of LOCALIO is to find those points where the latency saving justifies the code duplication.

When opening with AUTH_UNIX the code duplication to determine the correct credential is small and easy to review. If we ever wanted to support KRB5 or TLS I would be a lot less comfortable about reviewing the code duplication.

So I think it is worth considering whether an over-the-wire open is really all that costly. As I noted we already have an over-the-wire request at open time. We could conceivably send the LOCALIO-OPEN request at the same time so as not to add latency. We could receive the reply through the in-kernel backchannel so there is no RPC reply.

That might all be too complex and might not be justified. My point is that I think the trade-offs are subtle and I think the FAQ answer cuts off an avenue that hasn't really been explored.

Thanks,
NeilBrown

#48 lore raw

Re: [PATCH v13 19/19] nfs: add FAQ section to Documentation/filesystems/nfs/localio.rst

Quoted 32 lines · NeilBrown
Quoted 10 lines · Trond Myklebust
Quoted 9 lines
Quoted 8 lines · NeilBrown

While I'm not advocating for an over-the-wire request to map a filehandle to a struct nfsd_file*, I don't think you can convincingly
argue against it without concrete performance measurements.

Quoted 4 lines

What is the value of doing an open over the wire? What are you trying
to accomplish that can't be accomplished without going over the wire?

The advantage of going over the wire is avoiding code duplication. The cost is latency.  Obviously the goal of LOCALIO is to find those points where the latency saving justifies the code duplication.

When opening with AUTH_UNIX the code duplication to determine the correct credential is small and easy to review.  If we ever wanted to support KRB5 or TLS I would be a lot less comfortable about reviewing the code duplication.

So I think it is worth considering whether an over-the-wire open is really all that costly.  As I noted we already have an over-the-wire request at open time.  We could conceivably send the LOCALIO-OPEN request at the same time so as not to add latency.  We could receive the
reply through the in-kernel backchannel so there is no RPC reply.

Quoted 6 lines

That might all be too complex and might not be justified.  My point is
that I think the trade-offs are subtle and I think the FAQ answer cuts
off an avenue that hasn't really been explored.

So, your argument is that if there was a hypothetical situation where we wanted to add krb5 or TLS support, then we'd have more code to review?

The counter-argument would be that we've already established the right of the client to do I/O to the file. This will already have been done by an over-the-wire call to OPEN (NFSv4), ACCESS (NFSv3/NFSv4) or CREATE (NFSv3). Those calls will have used krb5 and/or TLS to authenticate the user. All that remains to be done is perform the I/O that was authorised by those calls.

Furthermore, we'd already have established that the client and the knfsd instance are running in the same kernel space on the same hardware (whether real or virtualised). There is no chance for a bad actor to compromise the one without also compromising the other. However, let's assume that somehow is possible: How does throwing in an on-the-wire protocol that is initiated by the one and interpreted by the other going to help, given that both have access to the exact same RPCSEC_GSS/TLS session and shared secret information via shared kernel memory?

So again, what problem are you trying to fix?

Signature · 3 lines
Trond Myklebust
Linux NFS client maintainer, Hammerspace
[email protected]
#49 lore raw

Re: [PATCH v13 19/19] nfs: add FAQ section to Documentation/filesystems/nfs/localio.rst

Quoted 44 lines · Trond Myklebust
Quoted 32 lines · NeilBrown
Quoted 10 lines · Trond Myklebust
Quoted 9 lines
Quoted 8 lines · NeilBrown

While I'm not advocating for an over-the-wire request to map a filehandle to a struct nfsd_file*, I don't think you can convincingly
argue against it without concrete performance measurements.

Quoted 4 lines

What is the value of doing an open over the wire? What are you trying
to accomplish that can't be accomplished without going over the wire?

The advantage of going over the wire is avoiding code duplication. The cost is latency.  Obviously the goal of LOCALIO is to find those points where the latency saving justifies the code duplication.

When opening with AUTH_UNIX the code duplication to determine the correct credential is small and easy to review.  If we ever wanted to support KRB5 or TLS I would be a lot less comfortable about reviewing the code duplication.

So I think it is worth considering whether an over-the-wire open is really all that costly.  As I noted we already have an over-the-wire request at open time.  We could conceivably send the LOCALIO-OPEN request at the same time so as not to add latency.  We could receive the
reply through the in-kernel backchannel so there is no RPC reply.

Quoted 6 lines

That might all be too complex and might not be justified.  My point is
that I think the trade-offs are subtle and I think the FAQ answer cuts
off an avenue that hasn't really been explored.

So, your argument is that if there was a hypothetical situation where we wanted to add krb5 or TLS support, then we'd have more code to review?

Quoted 6 lines

The counter-argument would be that we've already established the right of the client to do I/O to the file. This will already have been done by an over-the-wire call to OPEN (NFSv4), ACCESS (NFSv3/NFSv4) or CREATE (NFSv3). Those calls will have used krb5 and/or TLS to authenticate the user. All that remains to be done is perform the I/O that was authorised by those calls.

The other thing that remains is to get the correct 'struct cred *' to store in ->f_cred (or to use for lookup in the nfsd filecache).

Quoted 11 lines

Furthermore, we'd already have established that the client and the knfsd instance are running in the same kernel space on the same hardware (whether real or virtualised). There is no chance for a bad actor to compromise the one without also compromising the other. However, let's assume that somehow is possible: How does throwing in an on-the-wire protocol that is initiated by the one and interpreted by the other going to help, given that both have access to the exact same RPCSEC_GSS/TLS session and shared secret information via shared kernel memory?

Quoted 1 line

So again, what problem are you trying to fix?

Conversely: what exactly is this FAQ entry trying to argue against?

My current immediate goal is for the FAQ to be useful. It mostly is, but this one question/answer isn't clear to me.

Thanks,
NeilBrown

#50 lore raw

Re: [PATCH v13 19/19] nfs: add FAQ section to Documentation/filesystems/nfs/localio.rst

Quoted 66 lines · NeilBrown
Quoted 55 lines · Trond Myklebust
Quoted 41 lines · NeilBrown
Quoted 12 lines · Trond Myklebust
Quoted 11 lines
Quoted 10 lines

On Aug 25, 2024, at 9:56 PM, NeilBrown <[email protected]> wrote:

While I'm not advocating for an over-the-wire request to map a
filehandle to a struct nfsd_file*, I don't think you can convincingly
argue against it without concrete performance measurements.

Quoted 4 lines

What is the value of doing an open over the wire? What are you trying
to accomplish that can't be accomplished without going over the wire?

The advantage of going over the wire is avoiding code duplication.
The cost is latency.  Obviously the goal of LOCALIO is to find those
points where the latency saving justifies the code duplication.

When opening with AUTH_UNIX the code duplication to determine the correct credential is small and easy to review.  If we ever wanted to
support KRB5 or TLS I would be a lot less comfortable about reviewing
the code duplication.

So I think it is worth considering whether an over-the-wire open is
really all that costly.  As I noted we already have an over-the- wire
request at open time.  We could conceivably send the LOCALIO-OPEN request at the same time so as not to add latency.  We could receive
the
reply through the in-kernel backchannel so there is no RPC reply.

Quoted 7 lines

That might all be too complex and might not be justified.  My point
is
that I think the trade-offs are subtle and I think the FAQ answer cuts
off an avenue that hasn't really been explored.

So, your argument is that if there was a hypothetical situation where
we wanted to add krb5 or TLS support, then we'd have more code to review?

Quoted 9 lines

The counter-argument would be that we've already established the right
of the client to do I/O to the file. This will already have been done
by an over-the-wire call to OPEN (NFSv4), ACCESS (NFSv3/NFSv4) or CREATE (NFSv3). Those calls will have used krb5 and/or TLS to authenticate the user. All that remains to be done is perform the I/O
that was authorised by those calls.

Quoted 2 lines

The other thing that remains is to get the correct 'struct cred *' to store in ->f_cred (or to use for lookup in the nfsd filecache).

This was why the original code called up into the sunrpc server domain code, and hence did consult with mountd when needed. Is there any reason to believe that we shouldn't be able to do the same with future security models?
As I said, the client has direct access to all the RPCSEC_GSS/TLS session info, or other info that might be needed to look up the corresponding information in knfsd. In the worst case, we could fall back to sending on-the-wire info until the relevant context information has been re-established.

Quoted 17 lines
Quoted 17 lines

Furthermore, we'd already have established that the client and the knfsd instance are running in the same kernel space on the same hardware (whether real or virtualised). There is no chance for a bad
actor to compromise the one without also compromising the other. However, let's assume that somehow is possible: How does throwing in an
on-the-wire protocol that is initiated by the one and interpreted by
the other going to help, given that both have access to the exact same
RPCSEC_GSS/TLS session and shared secret information via shared kernel
memory?

Quoted 6 lines
Quoted 1 line

So again, what problem are you trying to fix?

Conversely:  what exactly is this FAQ entry trying to argue against?

My current immediate goal is for the FAQ to be useful.  It mostly is, but this one question/answer isn't clear to me.

The question arose from the feedback when Mike submitted the earlier drafts in the beginning of July. I was on vacation at the time, but my understanding is that several people (including some listed in the MAINTAINERS file) were asking questiona about how the code would support RPCSEC_GSS and TLS. The FAQ entry is a direct response to those questions.

I'm happy to ask Mike to drop that entry if everyone agrees that it is redundant, but the point is that at the time, there was a set of questions around this, and they were clearly blocking the ability to submit the code for merging.

Signature · 3 lines
Trond Myklebust
Linux NFS client maintainer, Hammerspace
[email protected]
#51 lore raw

Re: [PATCH v13 19/19] nfs: add FAQ section to Documentation/filesystems/nfs/localio.rst

In reply to NeilBrown
Quoted 66 lines · NeilBrown
Quoted 44 lines · Trond Myklebust
Quoted 32 lines · NeilBrown
Quoted 10 lines · Trond Myklebust
Quoted 9 lines
Quoted 8 lines · NeilBrown

While I'm not advocating for an over-the-wire request to map a filehandle to a struct nfsd_file*, I don't think you can convincingly
argue against it without concrete performance measurements.

Quoted 4 lines

What is the value of doing an open over the wire? What are you trying
to accomplish that can't be accomplished without going over the wire?

The advantage of going over the wire is avoiding code duplication. The cost is latency.  Obviously the goal of LOCALIO is to find those points where the latency saving justifies the code duplication.

When opening with AUTH_UNIX the code duplication to determine the correct credential is small and easy to review.  If we ever wanted to support KRB5 or TLS I would be a lot less comfortable about reviewing the code duplication.

So I think it is worth considering whether an over-the-wire open is really all that costly.  As I noted we already have an over-the-wire request at open time.  We could conceivably send the LOCALIO-OPEN request at the same time so as not to add latency.  We could receive the
reply through the in-kernel backchannel so there is no RPC reply.

Quoted 6 lines

That might all be too complex and might not be justified.  My point is
that I think the trade-offs are subtle and I think the FAQ answer cuts
off an avenue that hasn't really been explored.

So, your argument is that if there was a hypothetical situation where we wanted to add krb5 or TLS support, then we'd have more code to review?

Quoted 6 lines

The counter-argument would be that we've already established the right of the client to do I/O to the file. This will already have been done by an over-the-wire call to OPEN (NFSv4), ACCESS (NFSv3/NFSv4) or CREATE (NFSv3). Those calls will have used krb5 and/or TLS to authenticate the user. All that remains to be done is perform the I/O that was authorised by those calls.

The other thing that remains is to get the correct 'struct cred *' to store in ->f_cred (or to use for lookup in the nfsd filecache).

Quoted 11 lines

Furthermore, we'd already have established that the client and the knfsd instance are running in the same kernel space on the same hardware (whether real or virtualised). There is no chance for a bad actor to compromise the one without also compromising the other. However, let's assume that somehow is possible: How does throwing in an on-the-wire protocol that is initiated by the one and interpreted by the other going to help, given that both have access to the exact same RPCSEC_GSS/TLS session and shared secret information via shared kernel memory?

Quoted 6 lines
Quoted 1 line

So again, what problem are you trying to fix?

Conversely: what exactly is this FAQ entry trying to argue against?

My current immediate goal is for the FAQ to be useful. It mostly is, but this one question/answer isn't clear to me.

The current answer to question 6 isn't meant to be dealing in absolutes, nor does it have to (but I agree that "negating any benefit" should be softened given we don't _know_ how it'd play out without implementing open-over-the-wire entirely to benchmark).

We just need to give context for what motivated the current
implementation: network protocol avoidance where possible.

Given everything, do you have a suggestion for how to improve the answer to question 6? Happy to revise it however you like.

Here is the incremental patch I just came up with. Any better?

Documentation/filesystems/nfs/localio.rst +7 -6
index 4b6d63246479..5d652f637a97 100644
@@ -120,12 +120,13 @@ FAQ
    using RPC, beneficial?  Is the benefit pNFS specific?
 
    Avoiding the use of XDR and RPC for file opens is beneficial to
-   performance regardless of whether pNFS is used. However adding a
-   requirement to go over the wire to do an open and/or close ends up
-   negating any benefit of avoiding the wire for doing the I/O itself
-   when we’re dealing with small files. There is no benefit to replacing
-   the READ or WRITE with a new open and/or close operation that still
-   needs to go over the wire.
+   performance regardless of whether pNFS is used. Especially when
+   dealing with small files its best to avoid going over the wire
+   whenever possible, otherwise it could reduce or even negate the
+   benefits of avoiding the wire for doing the small file I/O itself.
+   Given LOCALIO's requirements the current approach of having the
+   client perform a server-side file open, without using RPC, is ideal.
+   If in the future requirements change then we can adapt accordingly.
 
 7. Why is LOCALIO only supported with UNIX Authentication (AUTH_UNIX)?
 
#52 lore raw
In reply to Mike Snitzer
Quoted 95 lines · Mike Snitzer

These latest changes are available in my git tree here:
https://git.kernel.org/pub/scm/linux/kernel/git/snitzer/linux.git/log/?h=nfs-localio-for-next

Changes since v12:
- Rebased to rearrange code to avoid "churn" that Jeff Layton felt distracting (biggest improvement came from folding switch from struct file to nfsd_file changes in from the start of the series)
- Updated relevant patch headers accordingly.
- Updated localio.rst to provide more performance data.
- Dropped v12's buggy "nfsd: fix nfsfh tracepoints to properly handle NULL rqstp" patch -- fixing localio to support fh_verify tracepoints will need more think-time and discussion, but they aren't of critical importance so fixing them doesn't need to hold things up.

Please also see v12's patch header for v12's extensive changes:
https://marc.info/?l=linux-nfs&m=172409139907981&w=2

Thanks,
Mike

Mike Snitzer (9):
nfs_common: factor out nfs_errtbl and nfs_stat_to_errno
nfs_common: factor out nfs4_errtbl and nfs4_stat_to_errno
nfs: factor out {encode,decode}_opaque_fixed to nfs_xdr.h
SUNRPC: remove call_allocate() BUG_ONs
nfs_common: add NFS LOCALIO auxiliary protocol enablement
nfsd: implement server support for NFS_LOCALIO_PROGRAM
nfs: pass struct nfsd_file to nfs_init_pgio and nfs_init_commit
nfs: implement client support for NFS_LOCALIO_PROGRAM
nfs: add Documentation/filesystems/nfs/localio.rst

NeilBrown (3):
nfsd: factor out __fh_verify to allow NULL rqstp to be passed
nfsd: add nfsd_file_acquire_local()
SUNRPC: replace program list with program array

Trond Myklebust (4):
nfs: enable localio for non-pNFS IO
pnfs/flexfiles: enable localio support
nfs/localio: use dedicated workqueues for filesystem read and write
nfs: add FAQ section to Documentation/filesystems/nfs/localio.rst

Weston Andros Adamson (3):
SUNRPC: add rpcauth_map_clnt_to_svc_cred_local
nfsd: add localio support
nfs: add localio support

 Documentation/filesystems/nfs/localio.rst | 276 ++++++++
 fs/Kconfig                                |   3 +
 fs/nfs/Kconfig                            |  15 +
 fs/nfs/Makefile                           |   1 +
 fs/nfs/client.c                           |  15 +-
 fs/nfs/filelayout/filelayout.c            |   6 +-
 fs/nfs/flexfilelayout/flexfilelayout.c    | 142 +++-
 fs/nfs/flexfilelayout/flexfilelayout.h    |   2 +
 fs/nfs/flexfilelayout/flexfilelayoutdev.c |   6 +
 fs/nfs/inode.c                            |  57 +-
 fs/nfs/internal.h                         |  61 +-
 fs/nfs/localio.c                          | 784 ++++++++++++++++++++++
 fs/nfs/nfs2xdr.c                          |  70 +-
 fs/nfs/nfs3xdr.c                          | 108 +--
 fs/nfs/nfs4xdr.c                          |  84 +--
 fs/nfs/nfstrace.h                         |  61 ++
 fs/nfs/pagelist.c                         |  16 +-
 fs/nfs/pnfs_nfs.c                         |   2 +-
 fs/nfs/write.c                            |  12 +-
 fs/nfs_common/Makefile                    |   5 +
 fs/nfs_common/common.c                    | 134 ++++
 fs/nfs_common/nfslocalio.c                | 194 ++++++
 fs/nfsd/Kconfig                           |  15 +
 fs/nfsd/Makefile                          |   1 +
 fs/nfsd/export.c                          |   8 +-
 fs/nfsd/filecache.c                       |  90 ++-
 fs/nfsd/filecache.h                       |   5 +
 fs/nfsd/localio.c                         | 183 +++++
 fs/nfsd/netns.h                           |   8 +-
 fs/nfsd/nfsctl.c                          |   2 +-
 fs/nfsd/nfsd.h                            |   6 +-
 fs/nfsd/nfsfh.c                           | 122 ++--
 fs/nfsd/nfsfh.h                           |   5 +
 fs/nfsd/nfssvc.c                          | 100 ++-
 fs/nfsd/trace.h                           |   3 +-
 fs/nfsd/vfs.h                             |  10 +
 include/linux/nfs.h                       |   9 +
 include/linux/nfs_common.h                |  17 +
 include/linux/nfs_fs_sb.h                 |  10 +
 include/linux/nfs_xdr.h                   |  20 +-
 include/linux/nfslocalio.h                |  56 ++
 include/linux/sunrpc/auth.h               |   4 +
 include/linux/sunrpc/svc.h                |   7 +-
 net/sunrpc/auth.c                         |  22 +
 net/sunrpc/clnt.c                         |   6 -
 net/sunrpc/svc.c                          |  68 +-
 net/sunrpc/svc_xprt.c                     |   2 +-
 net/sunrpc/svcauth_unix.c                 |   3 +-
Quoted 12 lines
 48 files changed, 2434 insertions(+), 402 deletions(-)
 create mode 100644 Documentation/filesystems/nfs/localio.rst
 create mode 100644 fs/nfs/localio.c
 create mode 100644 fs/nfs_common/common.c
 create mode 100644 fs/nfs_common/nfslocalio.c
 create mode 100644 fs/nfsd/localio.c
 create mode 100644 include/linux/nfs_common.h
 create mode 100644 include/linux/nfslocalio.h
Signature · 1 line
2.44.0

I've looked over the NFSD changes in this patch set. Some of them look ready to go (I've Acked those), others still need a little TLC.

For me, the big picture questions that remain are:

* How do we ensure that export options that modify RPC user translation (eg, all_squash) continue to work as expected on LOCALIO mounts?

* How do we ensure that a "stolen" FH can't be exploited by an NFS client running in a local container that does not have access to an export (server should return NFSERR_STALE)?

* Has LOCALIO been tested with two or more NFS servers running on the local system? One assumes they would be in separate containers.

The scenario I'm thinking of is two tenants, Pepsi and Coke, running in containers on the same physical host. They might want their own separate NFS servers, or they might share one server with appropriate access controls on the exports. Both of these scenarios need to ensure that Pepsi and Coke cannot open each other's files via LOCALIO.

Signature · 1 line
Chuck Lever
#53 lore raw
Quoted 126 lines · Chuck Lever
Quoted 95 lines · Mike Snitzer

These latest changes are available in my git tree here:
https://git.kernel.org/pub/scm/linux/kernel/git/snitzer/linux.git/log/?h=nfs-localio-for-next

Changes since v12:
- Rebased to rearrange code to avoid "churn" that Jeff Layton felt distracting (biggest improvement came from folding switch from struct file to nfsd_file changes in from the start of the series)
- Updated relevant patch headers accordingly.
- Updated localio.rst to provide more performance data.
- Dropped v12's buggy "nfsd: fix nfsfh tracepoints to properly handle NULL rqstp" patch -- fixing localio to support fh_verify tracepoints will need more think-time and discussion, but they aren't of critical importance so fixing them doesn't need to hold things up.

Please also see v12's patch header for v12's extensive changes:
https://marc.info/?l=linux-nfs&m=172409139907981&w=2

Thanks,
Mike

Mike Snitzer (9):
nfs_common: factor out nfs_errtbl and nfs_stat_to_errno
nfs_common: factor out nfs4_errtbl and nfs4_stat_to_errno
nfs: factor out {encode,decode}_opaque_fixed to nfs_xdr.h
SUNRPC: remove call_allocate() BUG_ONs
nfs_common: add NFS LOCALIO auxiliary protocol enablement
nfsd: implement server support for NFS_LOCALIO_PROGRAM
nfs: pass struct nfsd_file to nfs_init_pgio and nfs_init_commit
nfs: implement client support for NFS_LOCALIO_PROGRAM
nfs: add Documentation/filesystems/nfs/localio.rst

NeilBrown (3):
nfsd: factor out __fh_verify to allow NULL rqstp to be passed
nfsd: add nfsd_file_acquire_local()
SUNRPC: replace program list with program array

Trond Myklebust (4):
nfs: enable localio for non-pNFS IO
pnfs/flexfiles: enable localio support
nfs/localio: use dedicated workqueues for filesystem read and write
nfs: add FAQ section to Documentation/filesystems/nfs/localio.rst

Weston Andros Adamson (3):
SUNRPC: add rpcauth_map_clnt_to_svc_cred_local
nfsd: add localio support
nfs: add localio support

 Documentation/filesystems/nfs/localio.rst | 276 ++++++++
 fs/Kconfig                                |   3 +
 fs/nfs/Kconfig                            |  15 +
 fs/nfs/Makefile                           |   1 +
 fs/nfs/client.c                           |  15 +-
 fs/nfs/filelayout/filelayout.c            |   6 +-
 fs/nfs/flexfilelayout/flexfilelayout.c    | 142 +++-
 fs/nfs/flexfilelayout/flexfilelayout.h    |   2 +
 fs/nfs/flexfilelayout/flexfilelayoutdev.c |   6 +
 fs/nfs/inode.c                            |  57 +-
 fs/nfs/internal.h                         |  61 +-
 fs/nfs/localio.c                          | 784 ++++++++++++++++++++++
 fs/nfs/nfs2xdr.c                          |  70 +-
 fs/nfs/nfs3xdr.c                          | 108 +--
 fs/nfs/nfs4xdr.c                          |  84 +--
 fs/nfs/nfstrace.h                         |  61 ++
 fs/nfs/pagelist.c                         |  16 +-
 fs/nfs/pnfs_nfs.c                         |   2 +-
 fs/nfs/write.c                            |  12 +-
 fs/nfs_common/Makefile                    |   5 +
 fs/nfs_common/common.c                    | 134 ++++
 fs/nfs_common/nfslocalio.c                | 194 ++++++
 fs/nfsd/Kconfig                           |  15 +
 fs/nfsd/Makefile                          |   1 +
 fs/nfsd/export.c                          |   8 +-
 fs/nfsd/filecache.c                       |  90 ++-
 fs/nfsd/filecache.h                       |   5 +
 fs/nfsd/localio.c                         | 183 +++++
 fs/nfsd/netns.h                           |   8 +-
 fs/nfsd/nfsctl.c                          |   2 +-
 fs/nfsd/nfsd.h                            |   6 +-
 fs/nfsd/nfsfh.c                           | 122 ++--
 fs/nfsd/nfsfh.h                           |   5 +
 fs/nfsd/nfssvc.c                          | 100 ++-
 fs/nfsd/trace.h                           |   3 +-
 fs/nfsd/vfs.h                             |  10 +
 include/linux/nfs.h                       |   9 +
 include/linux/nfs_common.h                |  17 +
 include/linux/nfs_fs_sb.h                 |  10 +
 include/linux/nfs_xdr.h                   |  20 +-
 include/linux/nfslocalio.h                |  56 ++
 include/linux/sunrpc/auth.h               |   4 +
 include/linux/sunrpc/svc.h                |   7 +-
 net/sunrpc/auth.c                         |  22 +
 net/sunrpc/clnt.c                         |   6 -
 net/sunrpc/svc.c                          |  68 +-
 net/sunrpc/svc_xprt.c                     |   2 +-
 net/sunrpc/svcauth_unix.c                 |   3 +-
Quoted 12 lines
 48 files changed, 2434 insertions(+), 402 deletions(-)
 create mode 100644 Documentation/filesystems/nfs/localio.rst
 create mode 100644 fs/nfs/localio.c
 create mode 100644 fs/nfs_common/common.c
 create mode 100644 fs/nfs_common/nfslocalio.c
 create mode 100644 fs/nfsd/localio.c
 create mode 100644 include/linux/nfs_common.h
 create mode 100644 include/linux/nfslocalio.h
Signature · 1 line
2.44.0

I've looked over the NFSD changes in this patch set. Some of them look ready to go (I've Acked those), others still need a little TLC.

For me, the big picture questions that remain are:

* How do we ensure that export options that modify RPC user translation (eg, all_squash) continue to work as expected on LOCALIO mounts?

* How do we ensure that a "stolen" FH can't be exploited by an NFS client running in a local container that does not have access to an export (server should return NFSERR_STALE)?

* Has LOCALIO been tested with two or more NFS servers running on the local system? One assumes they would be in separate containers.

Quoted 6 lines

The scenario I'm thinking of is two tenants, Pepsi and Coke, running in containers on the same physical host. They might want their own separate NFS servers, or they might share one server with appropriate access controls on the exports. Both of these scenarios need to ensure that Pepsi and Coke cannot open each other's files via LOCALIO.

Each server instance is isolated in their respective nfsd_net. And localio was made to build on that isolation. So all creds used, etc are specific to each nfsd instance.

I haven't verified these scenarios "just work" as expected but localio preserves all the same checking and upcalls of standard NFS uses.

Mike

#54 lore raw
In reply to Mike Snitzer
Quoted 13 lines · Mike Snitzer

These latest changes are available in my git tree here:
https://git.kernel.org/pub/scm/linux/kernel/git/snitzer/linux.git/log/?h=nfs-localio-for-next

Changes since v12:
- Rebased to rearrange code to avoid "churn" that Jeff Layton felt distracting (biggest improvement came from folding switch from struct file to nfsd_file changes in from the start of the series)
- Updated relevant patch headers accordingly.
- Updated localio.rst to provide more performance data.
- Dropped v12's buggy "nfsd: fix nfsfh tracepoints to properly handle NULL rqstp" patch -- fixing localio to support fh_verify tracepoints will need more think-time and discussion, but they aren't of critical importance so fixing them doesn't need to hold things up.

Thanks for continuing to revise this. While I think there are still issues that need addressing, they are much smaller than some of the things we had to address in the past. It is certainly getting closer.

NeilBrown