diff mbox series

[v2,05/10] nfs: Add export support for weak cache consistency attributes

Message ID 20211219013803.324724-6-trondmy@kernel.org (mailing list archive)
State New, archived
Headers show
Series Assorted patches for knfsd | expand

Commit Message

Trond Myklebust Dec. 19, 2021, 1:37 a.m. UTC 
From: Trond Myklebust <trond.myklebust@primarydata.com>

Allow knfsd to request weak cache consistency attributes on files that
have delegations and/or have up to date attribute caches by propagating
the information to NFS that the attributes being requested are optional.

Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: Lance Shelton <lance.shelton@hammerspace.com>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/export.c          | 24 ++++++++++++++++++++++++
 fs/nfsd/nfs3xdr.c        |  8 ++++++--
 fs/nfsd/nfs4xdr.c        |  6 +++---
 fs/nfsd/vfs.c            | 14 ++++++++++++++
 fs/nfsd/vfs.h            |  5 +++--
 include/linux/exportfs.h |  3 +++
 6 files changed, 53 insertions(+), 7 deletions(-)

Comments

Daire Byrne Jan. 5, 2022, 4:10 p.m. UTC | #1 
Hi,

I was interested in testing these patches with our re-export
workloads. However, it looks like this specific patch and the previous
one (nfsd: Distinguish between required and optional NFSv3 post-op
attributes) are breaking re-exports in such a way as to cause
applications to randomly crash out with sigbus errors.

At a guess, loading applications and memory mapping data files are
particularly good at triggering this. I can see no errors logged on
either the re-export server or the eventual client (e.g. stale
filehandles). We mostly re-export NFSv4.2 servers to NFSv3 clients but
we do also have a few NFSv3 servers re-exported as NFSv3  too
(Netapps).

This only happens with patches 4 + 5 and vanilla 5.16-rc6 does not
have any issues. I haven't had a chance to dig much deeper yet, but
thought I'd flag it anyway.

Daire

On Sun, 19 Dec 2021 at 01:44, <trondmy@kernel.org> wrote:
>
> From: Trond Myklebust <trond.myklebust@primarydata.com>
>
> Allow knfsd to request weak cache consistency attributes on files that
> have delegations and/or have up to date attribute caches by propagating
> the information to NFS that the attributes being requested are optional.
diff mbox series

Patch

diff --git a/fs/nfs/export.c b/fs/nfs/export.c
index 171c424cb6d5..967f8902c49b 100644
--- a/fs/nfs/export.c
+++ b/fs/nfs/export.c
@@ -151,10 +151,34 @@  static u64 nfs_fetch_iversion(struct inode *inode)
 	return inode_peek_iversion_raw(inode);
 }
 
+static int nfs_exp_getattr(struct path *path, struct kstat *stat, bool force)
+{
+	const unsigned long check_valid =
+		NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_ATIME |
+		NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME |
+		NFS_INO_INVALID_SIZE | /* NFS_INO_INVALID_BLOCKS | */
+		NFS_INO_INVALID_OTHER | NFS_INO_INVALID_NLINK;
+	struct inode *inode = d_inode(path->dentry);
+	int flags = force ? AT_STATX_SYNC_AS_STAT : AT_STATX_DONT_SYNC;
+	int ret, ret2 = 0;
+
+	if (!force && nfs_check_cache_invalid(inode, check_valid))
+		ret2 = -EAGAIN;
+	ret = vfs_getattr(path, stat, STATX_BASIC_STATS & ~STATX_BLOCKS, flags);
+	if (ret < 0)
+		return ret;
+	stat->blocks = nfs_calc_block_size(stat->size);
+	if (S_ISDIR(inode->i_mode))
+		stat->blksize = NFS_SERVER(inode)->dtsize;
+	stat->result_mask |= STATX_BLOCKS;
+	return ret2;
+}
+
 const struct export_operations nfs_export_ops = {
 	.encode_fh = nfs_encode_fh,
 	.fh_to_dentry = nfs_fh_to_dentry,
 	.get_parent = nfs_get_parent,
+	.getattr = nfs_exp_getattr,
 	.fetch_iversion = nfs_fetch_iversion,
 	.flags = EXPORT_OP_NOWCC|EXPORT_OP_NOSUBTREECHK|
 		EXPORT_OP_CLOSE_BEFORE_UNLINK|EXPORT_OP_REMOTE_FS|
diff --git a/fs/nfsd/nfs3xdr.c b/fs/nfsd/nfs3xdr.c
index 6adfc40722fa..df6e29796494 100644
--- a/fs/nfsd/nfs3xdr.c
+++ b/fs/nfsd/nfs3xdr.c
@@ -420,6 +420,9 @@  __svcxdr_encode_post_op_attr(struct svc_rqst *rqstp, struct xdr_stream *xdr,
 			     const struct svc_fh *fhp, bool force)
 {
 	struct dentry *dentry = fhp->fh_dentry;
+	struct path path = {
+		.dentry = dentry,
+	};
 	struct kstat stat;
 
 	/*
@@ -427,9 +430,10 @@  __svcxdr_encode_post_op_attr(struct svc_rqst *rqstp, struct xdr_stream *xdr,
 	 * stale file handle. In this case, no attributes are
 	 * returned.
 	 */
-	if (!force || !dentry || !d_really_is_positive(dentry))
+	if (!dentry || !d_really_is_positive(dentry))
 		goto no_post_op_attrs;
-	if (fh_getattr(fhp, &stat) != nfs_ok)
+	path.mnt = fhp->fh_export->ex_path.mnt;
+	if (nfsd_getattr(&path, &stat, force) != nfs_ok)
 		goto no_post_op_attrs;
 
 	if (xdr_stream_encode_item_present(xdr) < 0)
diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
index 5a93a5db4fb0..8026925c121f 100644
--- a/fs/nfsd/nfs4xdr.c
+++ b/fs/nfsd/nfs4xdr.c
@@ -2862,9 +2862,9 @@  nfsd4_encode_fattr(struct xdr_stream *xdr, struct svc_fh *fhp,
 			goto out;
 	}
 
-	err = vfs_getattr(&path, &stat, STATX_BASIC_STATS, AT_STATX_SYNC_AS_STAT);
-	if (err)
-		goto out_nfserr;
+	status = nfsd_getattr(&path, &stat, true);
+	if (status)
+		goto out;
 	if ((bmval0 & (FATTR4_WORD0_FILES_AVAIL | FATTR4_WORD0_FILES_FREE |
 			FATTR4_WORD0_FILES_TOTAL | FATTR4_WORD0_MAXNAME)) ||
 	    (bmval1 & (FATTR4_WORD1_SPACE_AVAIL | FATTR4_WORD1_SPACE_FREE |
diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
index 0faa3839ea6c..eb9818432149 100644
--- a/fs/nfsd/vfs.c
+++ b/fs/nfsd/vfs.c
@@ -2411,3 +2411,17 @@  nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp,
 
 	return err? nfserrno(err) : 0;
 }
+
+
+__be32
+nfsd_getattr(struct path *p, struct kstat *stat, bool force)
+{
+	const struct export_operations *ops = p->dentry->d_sb->s_export_op;
+	int err;
+
+	if (ops->getattr)
+		err = ops->getattr(p, stat, force);
+	else
+		err = vfs_getattr(p, stat, STATX_BASIC_STATS, AT_STATX_SYNC_AS_STAT);
+	return err ? nfserrno(err) : 0;
+}
diff --git a/fs/nfsd/vfs.h b/fs/nfsd/vfs.h
index b21b76e6b9a8..6edae1b9a96e 100644
--- a/fs/nfsd/vfs.h
+++ b/fs/nfsd/vfs.h
@@ -132,6 +132,8 @@  __be32		nfsd_statfs(struct svc_rqst *, struct svc_fh *,
 __be32		nfsd_permission(struct svc_rqst *, struct svc_export *,
 				struct dentry *, int);
 
+__be32		nfsd_getattr(struct path *p, struct kstat *, bool);
+
 static inline int fh_want_write(struct svc_fh *fh)
 {
 	int ret;
@@ -156,8 +158,7 @@  static inline __be32 fh_getattr(const struct svc_fh *fh, struct kstat *stat)
 {
 	struct path p = {.mnt = fh->fh_export->ex_path.mnt,
 			 .dentry = fh->fh_dentry};
-	return nfserrno(vfs_getattr(&p, stat, STATX_BASIC_STATS,
-				    AT_STATX_SYNC_AS_STAT));
+	return nfsd_getattr(&p, stat, true);
 }
 
 static inline int nfsd_create_is_exclusive(int createmode)
diff --git a/include/linux/exportfs.h b/include/linux/exportfs.h
index 3260fe714846..58f36022787e 100644
--- a/include/linux/exportfs.h
+++ b/include/linux/exportfs.h
@@ -10,6 +10,8 @@  struct inode;
 struct iomap;
 struct super_block;
 struct vfsmount;
+struct path;
+struct kstat;
 
 /* limit the handle size to NFSv4 handle size now */
 #define MAX_HANDLE_SZ 128
@@ -224,6 +226,7 @@  struct export_operations {
 #define EXPORT_OP_SYNC_LOCKS		(0x20) /* Filesystem can't do
 						  asychronous blocking locks */
 	unsigned long	flags;
+	int (*getattr)(struct path *, struct kstat *, bool);
 };
 
 extern int exportfs_encode_inode_fh(struct inode *inode, struct fid *fid,