From: Chuck Lever <cel@kernel.org>
To: NeilBrown <neil@brown.name>, Jeff Layton <jlayton@kernel.org>,
Olga Kornievskaia <okorniev@redhat.com>,
Dai Ngo <dai.ngo@oracle.com>, Tom Talpey <tom@talpey.com>
Cc: <linux-nfs@vger.kernel.org>
Subject: [PATCH v1 08/27] NFSD: Use xdrgen XDR functions for NFSv2 SETATTR procedure
Date: Wed, 16 Sep 2026 12:28:39 -0400 [thread overview]
Message-ID: <20260916162859.2051-11-cel@kernel.org> (raw)
In-Reply-To: <20260916162859.2051-1-cel@kernel.org>
Replace the NFSPROC_SETATTR entry in the nfsd_procedures2 array
with an entry that dispatches the xdrgen-generated
nfs_svc_decode_sattrargs and nfs_svc_encode_attrstat. Wrapper
structures bridge the generated xdrgen types and the legacy
svc_fh, iattr, and kstat representations the NFSD VFS layer still
uses. The result reuses the attrstat wrapper from the GETATTR
conversion; the new nfsd_sattr_to_iattr() and
nfsd_timeval_to_timespec64() helpers translate the decoded sattr
into the iattr that nfsd_setattr() consumes.
The hand-coded svcxdr_decode_sattr() rejects a time-useconds value
greater than one second before converting it, preventing the
microsecond-to-nanosecond multiplication from wrapping to a valid
but incorrect value on 32-bit platforms. nfsd_sattr_to_iattr()
preserves that range check. Because it runs after the generated
decoder rather than inside it, an out-of-range value fails the
procedure with nfserr_io rather than the RPC GARBAGE_ARGS the
hand-coded decoder returned.
The pc_argzero field is now set to zero for the NFSv2 SETATTR
procedure. The xdrgen decoders are trusted to initialize all
arguments in the argp->xdrgen field, making the early defensive
memset unnecessary. The remaining argp fields are cleared as
needed. The wrapper's iattr is one of them: nfsd_sattr_to_iattr()
zeroes the whole structure rather than only ia_valid, because the
nfsd_vfs_setattr tracepoint records ia_size, ia_uid, and ia_gid
whether or not their ATTR bits are set.
This refactor replaces the use of svcxdr_encode_fattr(), so the
reference to the file handle can be released directly by
nfsd_proc_setattr(). A separate ->pc_release callback is thus no
longer needed. This makes it straightforward to confirm that the
file handle is properly released during every error flow.
Signed-off-by: Chuck Lever <cel@kernel.org>
---
fs/nfsd/nfsproc.c | 165 ++++++++++++++++++++++++++++++++++++----------
fs/nfsd/nfsxdr.c | 9 ---
fs/nfsd/xdr.h | 6 --
3 files changed, 131 insertions(+), 49 deletions(-)
diff --git a/fs/nfsd/nfsproc.c b/fs/nfsd/nfsproc.c
index 0196a554de5e..7fb5895f24d5 100644
--- a/fs/nfsd/nfsproc.c
+++ b/fs/nfsd/nfsproc.c
@@ -40,6 +40,14 @@ struct attrstat_wrapper {
static_assert(offsetof(struct attrstat_wrapper, xdrgen) == 0);
+struct sattrargs_wrapper {
+ struct sattrargs xdrgen;
+ struct svc_fh fh;
+ struct iattr iattrs;
+};
+
+static_assert(offsetof(struct sattrargs_wrapper, xdrgen) == 0);
+
static __be32 nfsd_map_status(__be32 status)
{
switch (status) {
@@ -81,6 +89,13 @@ nfsd_timespec64_to_timeval(struct timeval *dst,
dst->useconds = src->tv_nsec / NSEC_PER_USEC;
}
+static __always_inline void
+nfsd_timeval_to_timespec64(struct timespec64 *dst, const struct timeval *src)
+{
+ dst->tv_sec = src->seconds;
+ dst->tv_nsec = src->useconds * NSEC_PER_USEC;
+}
+
static u32
nfsd_mode_to_ftype(umode_t mode)
{
@@ -162,6 +177,76 @@ static __be32 nfsd_map_io_status(__be32 status)
return nfsd_map_status(status);
}
+/*
+ * Sun convention: a sattr time-useconds field of one full second (an
+ * otherwise out-of-range value) means "set this time to the current
+ * server time." It's needed to make permissions checks for the "touch"
+ * program across NFSv2 mounts work correctly. See description of
+ * sattr in section 6.1 of "NFS Illustrated" by Brent Callaghan,
+ * Addison-Wesley, ISBN 0-201-32750-5
+ */
+#define NFS2_SATTR_SET_TO_SERVER_TIME (1000000)
+
+static bool
+nfsd_sattr_to_iattr(struct svc_rqst *rqstp, struct iattr *iap,
+ const struct sattr *sattr)
+{
+ static const unsigned int dont_set_it = (unsigned int)-1;
+
+ /* Consumers read ia_size, ia_uid, and ia_gid unguarded by ia_valid */
+ memset(iap, 0, sizeof(*iap));
+
+ /*
+ * Some Sun NFS clients put 0xffff in the mode field when they
+ * mean 0xffffffff.
+ */
+ if (sattr->mode != dont_set_it && sattr->mode != 0xffff) {
+ iap->ia_valid |= ATTR_MODE;
+ iap->ia_mode = sattr->mode;
+ }
+ if (sattr->uid != dont_set_it) {
+ iap->ia_uid = make_kuid(nfsd_user_namespace(rqstp), sattr->uid);
+ if (uid_valid(iap->ia_uid))
+ iap->ia_valid |= ATTR_UID;
+ }
+ if (sattr->gid != dont_set_it) {
+ iap->ia_gid = make_kgid(nfsd_user_namespace(rqstp), sattr->gid);
+ if (gid_valid(iap->ia_gid))
+ iap->ia_valid |= ATTR_GID;
+ }
+ if (sattr->size != dont_set_it) {
+ iap->ia_valid |= ATTR_SIZE;
+ iap->ia_size = sattr->size;
+ }
+ if (sattr->atime.seconds != dont_set_it &&
+ sattr->atime.useconds != dont_set_it) {
+ /*
+ * Reject out-of-range useconds so the conversion to
+ * nanoseconds cannot wrap to a valid but incorrect
+ * value on 32-bit platforms.
+ */
+ if (sattr->atime.useconds > NFS2_SATTR_SET_TO_SERVER_TIME)
+ return false;
+ iap->ia_valid |= ATTR_ATIME | ATTR_ATIME_SET;
+ nfsd_timeval_to_timespec64(&iap->ia_atime, &sattr->atime);
+
+ if (sattr->atime.useconds == NFS2_SATTR_SET_TO_SERVER_TIME)
+ iap->ia_valid &= ~ATTR_ATIME_SET;
+ }
+ if (sattr->mtime.seconds != dont_set_it &&
+ sattr->mtime.useconds != dont_set_it) {
+ if (sattr->mtime.useconds > NFS2_SATTR_SET_TO_SERVER_TIME)
+ return false;
+ iap->ia_valid |= ATTR_MTIME | ATTR_MTIME_SET;
+ nfsd_timeval_to_timespec64(&iap->ia_mtime, &sattr->mtime);
+
+ if (sattr->mtime.useconds == NFS2_SATTR_SET_TO_SERVER_TIME)
+ iap->ia_valid &= ~(ATTR_ATIME_SET | ATTR_MTIME_SET);
+ }
+
+ return true;
+}
+
/*
* A full specification of each of the following NFSv2 procedures is
* available in RFC 1094 Section 2.2.
@@ -218,27 +303,32 @@ static __be32 nfsd_proc_getattr(struct svc_rqst *rqstp)
return rpc_success;
}
-/*
- * Set a file's attributes
- * N.B. After this call resp->fh needs an fh_put
+/**
+ * nfsd_proc_setattr - SETATTR: Set file attributes
+ * @rqstp: RPC transaction context
+ *
+ * Return:
+ * %rpc_success: RPC executed successfully
+ *
+ * RPC synopsis:
+ * attrstat NFSPROC_SETATTR(sattrargs) = 2;
*/
-static __be32
-nfsd_proc_setattr(struct svc_rqst *rqstp)
+static __be32 nfsd_proc_setattr(struct svc_rqst *rqstp)
{
- struct nfsd_sattrargs *argp = rqstp->rq_argp;
- struct nfsd_attrstat *resp = rqstp->rq_resp;
- struct iattr *iap = &argp->attrs;
- struct nfsd_attrs attrs = {
+ struct sattrargs_wrapper *argp = rqstp->rq_argp;
+ struct attrstat_wrapper *resp = rqstp->rq_resp;
+ struct kstat *statp = &resp->stat;
+ struct iattr *iap = &argp->iattrs;
+ struct nfsd_attrs nattrs = {
.na_iattr = iap,
};
- struct svc_fh *fhp;
- int hosterr;
+ struct svc_fh *fhp = &argp->fh;
- dprintk("nfsd: SETATTR %s, valid=%x, size=%ld\n",
- SVCFH_fmt(&argp->fh),
- argp->attrs.ia_valid, (long) argp->attrs.ia_size);
-
- fhp = fh_copy(&resp->fh, &argp->fh);
+ nfsd_fhandle_to_svc_fh(fhp, &argp->xdrgen.file);
+ if (!nfsd_sattr_to_iattr(rqstp, iap, &argp->xdrgen.attributes)) {
+ resp->xdrgen.status = nfserr_io;
+ goto out;
+ }
/*
* NFSv2 does not differentiate between "set-[ac]time-to-now"
@@ -255,6 +345,8 @@ nfsd_proc_setattr(struct svc_rqst *rqstp)
#define MAX_TOUCH_TIME_ERROR (30*60)
if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET &&
iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) {
+ int hosterr;
+
/*
* Looks probable.
*
@@ -264,13 +356,13 @@ nfsd_proc_setattr(struct svc_rqst *rqstp)
*/
time64_t delta = iap->ia_atime.tv_sec - ktime_get_real_seconds();
- resp->status = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP);
- if (resp->status != nfs_ok)
+ resp->xdrgen.status = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP);
+ if (resp->xdrgen.status != nfs_ok)
goto out;
hosterr = fh_want_write(fhp);
if (hosterr) {
- resp->status = nfserrno(hosterr);
+ resp->xdrgen.status = nfserrno(hosterr);
goto out;
}
@@ -287,13 +379,19 @@ nfsd_proc_setattr(struct svc_rqst *rqstp)
}
}
- resp->status = nfsd_setattr(rqstp, fhp, &attrs, NULL);
- if (resp->status != nfs_ok)
+ resp->xdrgen.status = nfsd_setattr(rqstp, fhp, &nattrs, NULL);
+ if (resp->xdrgen.status != nfs_ok)
goto out;
- resp->status = fh_getattr(&resp->fh, &resp->stat);
+ resp->xdrgen.status = fh_getattr(fhp, statp);
+
out:
- resp->status = nfsd_map_status(resp->status);
+ if (resp->xdrgen.status == nfs_ok)
+ nfsd_stat_to_fattr(rqstp, &resp->xdrgen.u.attributes, statp, fhp);
+ else
+ resp->xdrgen.status = nfsd_map_status(resp->xdrgen.status);
+
+ fh_put(fhp);
return rpc_success;
}
@@ -827,16 +925,15 @@ static const struct svc_procedure nfsd_procedures2[18] = {
.pc_name = "GETATTR",
},
[NFSPROC_SETATTR] = {
- .pc_func = nfsd_proc_setattr,
- .pc_decode = nfssvc_decode_sattrargs,
- .pc_encode = nfssvc_encode_attrstatres,
- .pc_release = nfssvc_release_attrstat,
- .pc_argsize = sizeof(struct nfsd_sattrargs),
- .pc_argzero = sizeof(struct nfsd_sattrargs),
- .pc_ressize = sizeof(struct nfsd_attrstat),
- .pc_cachetype = RC_REPLBUFF,
- .pc_xdrressize = ST+AT,
- .pc_name = "SETATTR",
+ .pc_func = nfsd_proc_setattr,
+ .pc_decode = nfs_svc_decode_sattrargs,
+ .pc_encode = nfs_svc_encode_attrstat,
+ .pc_argsize = sizeof(struct sattrargs_wrapper),
+ .pc_argzero = 0,
+ .pc_ressize = sizeof(struct attrstat_wrapper),
+ .pc_cachetype = RC_REPLBUFF,
+ .pc_xdrressize = NFS2_attrstat_sz,
+ .pc_name = "SETATTR",
},
[NFSPROC_ROOT] = {
.pc_func = nfsd_proc_root,
@@ -1014,7 +1111,7 @@ static const struct svc_procedure nfsd_procedures2[18] = {
*/
union nfsd_xdrstore {
struct fhandle_wrapper fhandle;
- struct nfsd_sattrargs sattr;
+ struct sattrargs_wrapper sattrargs;
struct nfsd_diropargs dirop;
struct nfsd_readargs read;
struct nfsd_writeargs write;
diff --git a/fs/nfsd/nfsxdr.c b/fs/nfsd/nfsxdr.c
index d55e7d6414a0..174e6f5622f9 100644
--- a/fs/nfsd/nfsxdr.c
+++ b/fs/nfsd/nfsxdr.c
@@ -302,15 +302,6 @@ nfssvc_decode_fhandleargs(struct svc_rqst *rqstp, struct xdr_stream *xdr)
return svcxdr_decode_fhandle(xdr, &args->fh);
}
-bool
-nfssvc_decode_sattrargs(struct svc_rqst *rqstp, struct xdr_stream *xdr)
-{
- struct nfsd_sattrargs *args = rqstp->rq_argp;
-
- return svcxdr_decode_fhandle(xdr, &args->fh) &&
- svcxdr_decode_sattr(rqstp, xdr, &args->attrs);
-}
-
bool
nfssvc_decode_diropargs(struct svc_rqst *rqstp, struct xdr_stream *xdr)
{
diff --git a/fs/nfsd/xdr.h b/fs/nfsd/xdr.h
index 05b3a18c879a..9983fe212f3a 100644
--- a/fs/nfsd/xdr.h
+++ b/fs/nfsd/xdr.h
@@ -8,11 +8,6 @@
#include "vfs.h"
-struct nfsd_sattrargs {
- struct svc_fh fh;
- struct iattr attrs;
-};
-
struct nfsd_diropargs {
struct svc_fh fh;
char * name;
@@ -120,7 +115,6 @@ struct nfsd_statfsres {
};
bool nfssvc_decode_fhandleargs(struct svc_rqst *rqstp, struct xdr_stream *xdr);
-bool nfssvc_decode_sattrargs(struct svc_rqst *rqstp, struct xdr_stream *xdr);
bool nfssvc_decode_diropargs(struct svc_rqst *rqstp, struct xdr_stream *xdr);
bool nfssvc_decode_readargs(struct svc_rqst *rqstp, struct xdr_stream *xdr);
bool nfssvc_decode_writeargs(struct svc_rqst *rqstp, struct xdr_stream *xdr);
--
2.55.0
next prev parent reply other threads:[~2026-09-16 16:29 UTC|newest]
Thread overview: 34+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-09-16 16:28 [PATCH v1 00/27] Convert server-side NFSv2 XDR to use xdrgen Chuck Lever
2026-09-16 16:28 ` [PATCH v1 01/27] Documentation: Add the RPC language description of NFSv2 Chuck Lever
2026-09-16 16:28 ` [PATCH v1 1/2] NFSD: map fh_verify() status codes for NFS_ACLv2 replies Chuck Lever
2026-09-17 11:57 ` Jeff Layton
2026-09-16 16:28 ` [PATCH v1 02/27] NFSD: Add infrastructure for generating NFSv2 XDR encoders and decoders Chuck Lever
2026-09-16 16:28 ` [PATCH v1 2/2] NFSD: map fh_verify() status codes for NFS_ACLv3 replies Chuck Lever
2026-09-17 11:57 ` Jeff Layton
2026-09-16 16:28 ` [PATCH v1 03/27] NFSD: Use xdrgen-generated NFSv2 protocol definitions Chuck Lever
2026-09-16 16:28 ` [PATCH v1 04/27] NFSD: Remove '#include "xdr.h"' from fs/nfsd/xdr3.h Chuck Lever
2026-09-16 16:28 ` [PATCH v1 05/27] NFSD: Relocate the NFSv2 XDR storage union into nfsproc.c Chuck Lever
2026-09-16 16:28 ` [PATCH v1 06/27] NFSD: Use xdrgen XDR functions for the NFSv2 NULL procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 07/27] NFSD: Use xdrgen XDR functions for NFSv2 GETATTR procedure Chuck Lever
2026-09-16 16:28 ` Chuck Lever [this message]
2026-09-16 16:28 ` [PATCH v1 09/27] NFSD: Use xdrgen XDR functions for the NFSv2 ROOT procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 10/27] NFSD: Use xdrgen XDR functions for the NFSv2 LOOKUP procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 11/27] NFSD: Use xdrgen XDR functions for NFSv2 READLINK procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 12/27] NFSD: Use xdrgen XDR functions for NFSv2 READ procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 13/27] NFSD: Use xdrgen XDR functions for the NFSv2 WRITECACHE procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 14/27] NFSD: Use xdrgen XDR functions for NFSv2 WRITE procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 15/27] NFSD: Refactor nfsd_proc_create() Chuck Lever
2026-09-16 16:28 ` [PATCH v1 16/27] NFSD: Use xdrgen XDR functions for NFSv2 CREATE procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 17/27] NFSD: Use xdrgen XDR functions for the NFSv2 REMOVE procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 18/27] NFSD: Use xdrgen XDR functions for the NFSv2 RENAME procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 19/27] NFSD: Use xdrgen XDR functions for the NFSv2 LINK procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 20/27] NFSD: Use xdrgen XDR functions for NFSv2 SYMLINK procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 21/27] NFSD: Use xdrgen XDR functions for NFSv2 MKDIR procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 22/27] NFSD: Use xdrgen XDR functions for NFSv2 RMDIR procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 23/27] NFSD: Use xdrgen XDR functions for the NFSv2 STATFS procedure Chuck Lever
2026-09-16 16:28 ` [PATCH v1 24/27] NFSD: Use xdrgen XDR functions for NFSv2 READDIR arguments Chuck Lever
2026-09-16 16:28 ` [PATCH v1 25/27] NFSD: Add a streaming directory reader Chuck Lever
2026-09-16 16:28 ` [PATCH v1 26/27] NFSD: Refactor NFSv2 directory cookie encoding Chuck Lever
2026-09-16 16:28 ` [PATCH v1 27/27] NFSD: Use xdrgen XDR functions for NFSv2 READDIR results Chuck Lever
2026-09-17 11:56 ` [PATCH v1 00/27] Convert server-side NFSv2 XDR to use xdrgen Jeff Layton
2026-09-17 14:57 ` Chuck Lever
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20260916162859.2051-11-cel@kernel.org \
--to=cel@kernel.org \
--cc=dai.ngo@oracle.com \
--cc=jlayton@kernel.org \
--cc=linux-nfs@vger.kernel.org \
--cc=neil@brown.name \
--cc=okorniev@redhat.com \
--cc=tom@talpey.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox