Merge tag 'nfsd-5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux

Pull nfsd updates from Chuck Lever:
 "We introduce 'courteous server' in this release. Previously NFSD would
  purge open and lock state for an unresponsive client after one lease
  period (typically 90 seconds). Now, after one lease period, another
  client can open and lock those files and the unresponsive client's
  lease is purged; otherwise if the unresponsive client's open and lock
  state is uncontended, the server retains that open and lock state for
  up to 24 hours, allowing the client's workload to resume after a
  lengthy network partition.

  A longstanding issue with NFSv4 file creation is also addressed.
  Previously a file creation can fail internally, returning an error to
  the client, but leave the newly created file in place as an artifact.
  The file creation code path has been reorganized so that internal
  failures and race conditions are less likely to result in an unwanted
  file creation.

  A fault injector has been added to help exercise paths that are run
  during kernel metadata cache invalidation. These caches contain
  information maintained by user space about exported filesystems. Many
  of our test workloads do not trigger cache invalidation.

  There is one patch that is needed to support PREEMPT_RT and a fix for
  an ancient 'sleep while spin-locked' splat that seems to have become
  easier to hit since v5.18-rc3"

* tag 'nfsd-5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux: (36 commits)
  NFSD: nfsd_file_put() can sleep
  NFSD: Add documenting comment for nfsd4_release_lockowner()
  NFSD: Modernize nfsd4_release_lockowner()
  NFSD: Fix possible sleep during nfsd4_release_lockowner()
  nfsd: destroy percpu stats counters after reply cache shutdown
  nfsd: Fix null-ptr-deref in nfsd_fill_super()
  nfsd: Unregister the cld notifier when laundry_wq create failed
  SUNRPC: Use RMW bitops in single-threaded hot paths
  NFSD: Clean up the show_nf_flags() macro
  NFSD: Trace filecache opens
  NFSD: Move documenting comment for nfsd4_process_open2()
  NFSD: Fix whitespace
  NFSD: Remove dprintk call sites from tail of nfsd4_open()
  NFSD: Instantiate a struct file when creating a regular NFSv4 file
  NFSD: Clean up nfsd_open_verified()
  NFSD: Remove do_nfsd_create()
  NFSD: Refactor NFSv4 OPEN(CREATE)
  NFSD: Refactor NFSv3 CREATE
  NFSD: Refactor nfsd_create_setattr()
  NFSD: Avoid calling fh_drop_write() twice in do_nfsd_create()
  ...

1 files changed, 55 insertions, 200 deletions

diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
index c22ad0532e8e..840e3af63a6f 100644
--- a/fs/nfsd/vfs.c
+++ b/fs/nfsd/vfs.c
@@ -827,14 +827,23 @@ retry:
 	return err;
 }
 
+/**
+ * nfsd_open_verified - Open a regular file for the filecache
+ * @rqstp: RPC request
+ * @fhp: NFS filehandle of the file to open
+ * @may_flags: internal permission flags
+ * @filp: OUT: open "struct file *"
+ *
+ * Returns an nfsstat value in network byte order.
+ */
 __be32
-nfsd_open_verified(struct svc_rqst *rqstp, struct svc_fh *fhp, umode_t type,
-		int may_flags, struct file **filp)
+nfsd_open_verified(struct svc_rqst *rqstp, struct svc_fh *fhp, int may_flags,
+		   struct file **filp)
 {
 	__be32 err;
 
 	validate_process_creds();
-	err = __nfsd_open(rqstp, fhp, type, may_flags, filp);
+	err = __nfsd_open(rqstp, fhp, S_IFREG, may_flags, filp);
 	validate_process_creds();
 	return err;
 }
@@ -849,17 +858,11 @@ nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
 		  struct splice_desc *sd)
 {
 	struct svc_rqst *rqstp = sd->u.data;
-	struct page **pp = rqstp->rq_next_page;
-	struct page *page = buf->page;
 
-	if (rqstp->rq_res.page_len == 0) {
-		svc_rqst_replace_page(rqstp, page);
+	svc_rqst_replace_page(rqstp, buf->page);
+	if (rqstp->rq_res.page_len == 0)
 		rqstp->rq_res.page_base = buf->offset;
-	} else if (page != pp[-1]) {
-		svc_rqst_replace_page(rqstp, page);
-	}
 	rqstp->rq_res.page_len += sd->len;
-
 	return sd->len;
 }
 
@@ -1187,14 +1190,26 @@ out:
 	return err;
 }
 
-static __be32
-nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp,
-			struct iattr *iap)
+/**
+ * nfsd_create_setattr - Set a created file's attributes
+ * @rqstp: RPC transaction being executed
+ * @fhp: NFS filehandle of parent directory
+ * @resfhp: NFS filehandle of new object
+ * @iap: requested attributes of new object
+ *
+ * Returns nfs_ok on success, or an nfsstat in network byte order.
+ */
+__be32
+nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp,
+		    struct svc_fh *resfhp, struct iattr *iap)
 {
+	__be32 status;
+
 	/*
-	 * Mode has already been set earlier in create:
+	 * Mode has already been set by file creation.
 	 */
 	iap->ia_valid &= ~ATTR_MODE;
+
 	/*
 	 * Setting uid/gid works only for root.  Irix appears to
 	 * send along the gid on create when it tries to implement
@@ -1202,10 +1217,31 @@ nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp,
 	 */
 	if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID))
 		iap->ia_valid &= ~(ATTR_UID|ATTR_GID);
+
+	/*
+	 * Callers expect new file metadata to be committed even
+	 * if the attributes have not changed.
+	 */
 	if (iap->ia_valid)
-		return nfsd_setattr(rqstp, resfhp, iap, 0, (time64_t)0);
-	/* Callers expect file metadata to be committed here */
-	return nfserrno(commit_metadata(resfhp));
+		status = nfsd_setattr(rqstp, resfhp, iap, 0, (time64_t)0);
+	else
+		status = nfserrno(commit_metadata(resfhp));
+
+	/*
+	 * Transactional filesystems had a chance to commit changes
+	 * for both parent and child simultaneously making the
+	 * following commit_metadata a noop in many cases.
+	 */
+	if (!status)
+		status = nfserrno(commit_metadata(fhp));
+
+	/*
+	 * Update the new filehandle to pick up the new attributes.
+	 */
+	if (!status)
+		status = fh_update(resfhp);
+
+	return status;
 }
 
 /* HPUX client sometimes creates a file in mode 000, and sets size to 0.
@@ -1232,7 +1268,6 @@ nfsd_create_locked(struct svc_rqst *rqstp, struct svc_fh *fhp,
 	struct dentry	*dentry, *dchild;
 	struct inode	*dirp;
 	__be32		err;
-	__be32		err2;
 	int		host_err;
 
 	dentry = fhp->fh_dentry;
@@ -1305,22 +1340,8 @@ nfsd_create_locked(struct svc_rqst *rqstp, struct svc_fh *fhp,
 	if (host_err < 0)
 		goto out_nfserr;
 
-	err = nfsd_create_setattr(rqstp, resfhp, iap);
+	err = nfsd_create_setattr(rqstp, fhp, resfhp, iap);
 
-	/*
-	 * nfsd_create_setattr already committed the child.  Transactional
-	 * filesystems had a chance to commit changes for both parent and
-	 * child simultaneously making the following commit_metadata a
-	 * noop.
-	 */
-	err2 = nfserrno(commit_metadata(fhp));
-	if (err2)
-		err = err2;
-	/*
-	 * Update the file handle to get the new inode info.
-	 */
-	if (!err)
-		err = fh_update(resfhp);
 out:
 	dput(dchild);
 	return err;
@@ -1376,172 +1397,6 @@ nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp,
 }
 
 /*
- * NFSv3 and NFSv4 version of nfsd_create
- */
-__be32
-do_nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp,
-		char *fname, int flen, struct iattr *iap,
-		struct svc_fh *resfhp, int createmode, u32 *verifier,
-	        bool *truncp, bool *created)
-{
-	struct dentry	*dentry, *dchild = NULL;
-	struct inode	*dirp;
-	__be32		err;
-	int		host_err;
-	__u32		v_mtime=0, v_atime=0;
-
-	err = nfserr_perm;
-	if (!flen)
-		goto out;
-	err = nfserr_exist;
-	if (isdotent(fname, flen))
-		goto out;
-	if (!(iap->ia_valid & ATTR_MODE))
-		iap->ia_mode = 0;
-	err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
-	if (err)
-		goto out;
-
-	dentry = fhp->fh_dentry;
-	dirp = d_inode(dentry);
-
-	host_err = fh_want_write(fhp);
-	if (host_err)
-		goto out_nfserr;
-
-	fh_lock_nested(fhp, I_MUTEX_PARENT);
-
-	/*
-	 * Compose the response file handle.
-	 */
-	dchild = lookup_one_len(fname, dentry, flen);
-	host_err = PTR_ERR(dchild);
-	if (IS_ERR(dchild))
-		goto out_nfserr;
-
-	/* If file doesn't exist, check for permissions to create one */
-	if (d_really_is_negative(dchild)) {
-		err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
-		if (err)
-			goto out;
-	}
-
-	err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
-	if (err)
-		goto out;
-
-	if (nfsd_create_is_exclusive(createmode)) {
-		/* solaris7 gets confused (bugid 4218508) if these have
-		 * the high bit set, as do xfs filesystems without the
-		 * "bigtime" feature.  So just clear the high bits. If this is
-		 * ever changed to use different attrs for storing the
-		 * verifier, then do_open_lookup() will also need to be fixed
-		 * accordingly.
-		 */
-		v_mtime = verifier[0]&0x7fffffff;
-		v_atime = verifier[1]&0x7fffffff;
-	}
-	
-	if (d_really_is_positive(dchild)) {
-		err = 0;
-
-		switch (createmode) {
-		case NFS3_CREATE_UNCHECKED:
-			if (! d_is_reg(dchild))
-				goto out;
-			else if (truncp) {
-				/* in nfsv4, we need to treat this case a little
-				 * differently.  we don't want to truncate the
-				 * file now; this would be wrong if the OPEN
-				 * fails for some other reason.  furthermore,
-				 * if the size is nonzero, we should ignore it
-				 * according to spec!
-				 */
-				*truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size;
-			}
-			else {
-				iap->ia_valid &= ATTR_SIZE;
-				goto set_attr;
-			}
-			break;
-		case NFS3_CREATE_EXCLUSIVE:
-			if (   d_inode(dchild)->i_mtime.tv_sec == v_mtime
-			    && d_inode(dchild)->i_atime.tv_sec == v_atime
-			    && d_inode(dchild)->i_size  == 0 ) {
-				if (created)
-					*created = true;
-				break;
-			}
-			fallthrough;
-		case NFS4_CREATE_EXCLUSIVE4_1:
-			if (   d_inode(dchild)->i_mtime.tv_sec == v_mtime
-			    && d_inode(dchild)->i_atime.tv_sec == v_atime
-			    && d_inode(dchild)->i_size  == 0 ) {
-				if (created)
-					*created = true;
-				goto set_attr;
-			}
-			fallthrough;
-		case NFS3_CREATE_GUARDED:
-			err = nfserr_exist;
-		}
-		fh_drop_write(fhp);
-		goto out;
-	}
-
-	if (!IS_POSIXACL(dirp))
-		iap->ia_mode &= ~current_umask();
-
-	host_err = vfs_create(&init_user_ns, dirp, dchild, iap->ia_mode, true);
-	if (host_err < 0) {
-		fh_drop_write(fhp);
-		goto out_nfserr;
-	}
-	if (created)
-		*created = true;
-
-	nfsd_check_ignore_resizing(iap);
-
-	if (nfsd_create_is_exclusive(createmode)) {
-		/* Cram the verifier into atime/mtime */
-		iap->ia_valid = ATTR_MTIME|ATTR_ATIME
-			| ATTR_MTIME_SET|ATTR_ATIME_SET;
-		/* XXX someone who knows this better please fix it for nsec */ 
-		iap->ia_mtime.tv_sec = v_mtime;
-		iap->ia_atime.tv_sec = v_atime;
-		iap->ia_mtime.tv_nsec = 0;
-		iap->ia_atime.tv_nsec = 0;
-	}
-
- set_attr:
-	err = nfsd_create_setattr(rqstp, resfhp, iap);
-
-	/*
-	 * nfsd_create_setattr already committed the child
-	 * (and possibly also the parent).
-	 */
-	if (!err)
-		err = nfserrno(commit_metadata(fhp));
-
-	/*
-	 * Update the filehandle to get the new inode info.
-	 */
-	if (!err)
-		err = fh_update(resfhp);
-
- out:
-	fh_unlock(fhp);
-	if (dchild && !IS_ERR(dchild))
-		dput(dchild);
-	fh_drop_write(fhp);
- 	return err;
- 
- out_nfserr:
-	err = nfserrno(host_err);
-	goto out;
-}
-
-/*
  * Read a symlink. On entry, *lenp must contain the maximum path length that
  * fits into the buffer. On return, it contains the true length.
  * N.B. After this call fhp needs an fh_put