2021-01-29 19:28:36

by J. Bruce Fields

Subject: [PATCH 2/2 v2] nfsd: skip some unnecessary stats in the v4 case

From: "J. Bruce Fields" <[email protected]>

In the typical case of v4 and an i_version-supporting filesystem, we can
skip a stat which is only required to fake up a change attribute from
ctime.

Signed-off-by: J. Bruce Fields <[email protected]>
---
fs/nfsd/nfs3xdr.c | 42 +++++++++++++++++++++++++-----------------
1 file changed, 25 insertions(+), 17 deletions(-)

diff --git a/fs/nfsd/nfs3xdr.c b/fs/nfsd/nfs3xdr.c
index 821db21ba072..8bcda5274089 100644
--- a/fs/nfsd/nfs3xdr.c
+++ b/fs/nfsd/nfs3xdr.c
@@ -252,6 +252,11 @@ encode_wcc_data(struct svc_rqst *rqstp, __be32 *p, struct svc_fh *fhp)
return encode_post_op_attr(rqstp, p, fhp);
}

+static bool fs_supports_change_attribute(struct super_block *sb)
+{
+ return sb->s_flags & SB_I_VERSION || sb->s_export_op->fetch_iversion;
+}
+
/*
* Fill in the pre_op attr for the wcc data
*/
@@ -260,24 +265,25 @@ void fill_pre_wcc(struct svc_fh *fhp)
struct inode *inode;
struct kstat stat;
bool v4 = (fhp->fh_maxsize == NFS4_FHSIZE);
- __be32 err;

if (fhp->fh_no_wcc || fhp->fh_pre_saved)
return;
inode = d_inode(fhp->fh_dentry);
- err = fh_getattr(fhp, &stat);
- if (err) {
- /* Grab the times from inode anyway */
- stat.mtime = inode->i_mtime;
- stat.ctime = inode->i_ctime;
- stat.size = inode->i_size;
+ if (fs_supports_change_attribute(inode->i_sb) || !v4) {
+ __be32 err = fh_getattr(fhp, &stat);
+ if (err) {
+ /* Grab the times from inode anyway */
+ stat.mtime = inode->i_mtime;
+ stat.ctime = inode->i_ctime;
+ stat.size = inode->i_size;
+ }
+ fhp->fh_pre_mtime = stat.mtime;
+ fhp->fh_pre_ctime = stat.ctime;
+ fhp->fh_pre_size = stat.size;
}
if (v4)
fhp->fh_pre_change = nfsd4_change_attribute(&stat, inode);

- fhp->fh_pre_mtime = stat.mtime;
- fhp->fh_pre_ctime = stat.ctime;
- fhp->fh_pre_size = stat.size;
fhp->fh_pre_saved = true;
}

@@ -288,7 +294,6 @@ void fill_post_wcc(struct svc_fh *fhp)
{
bool v4 = (fhp->fh_maxsize == NFS4_FHSIZE);
struct inode *inode = d_inode(fhp->fh_dentry);
- __be32 err;

if (fhp->fh_no_wcc)
return;
@@ -296,12 +301,15 @@ void fill_post_wcc(struct svc_fh *fhp)
if (fhp->fh_post_saved)
printk("nfsd: inode locked twice during operation.\n");

- err = fh_getattr(fhp, &fhp->fh_post_attr);
- if (err) {
- fhp->fh_post_saved = false;
- fhp->fh_post_attr.ctime = inode->i_ctime;
- } else
- fhp->fh_post_saved = true;
+ fhp->fh_post_saved = true;
+
+ if (fs_supports_change_attribute(inode->i_sb) || !v4) {
+ __be32 err = fh_getattr(fhp, &fhp->fh_post_attr);
+ if (err) {
+ fhp->fh_post_saved = false;
+ fhp->fh_post_attr.ctime = inode->i_ctime;
+ }
+ }
if (v4)
fhp->fh_post_change =
nfsd4_change_attribute(&fhp->fh_post_attr, inode);
--
2.29.2

2021-01-30 06:05:35

by Christoph Hellwig

Subject: Re: [PATCH 2/2 v2] nfsd: skip some unnecessary stats in the v4 case

On Fri, Jan 29, 2021 at 02:27:01PM -0500, J. Bruce Fields wrote:
> From: "J. Bruce Fields" <[email protected]>
>
> In the typical case of v4 and an i_version-supporting filesystem, we can
> skip a stat which is only required to fake up a change attribute from
> ctime.
>
> Signed-off-by: J. Bruce Fields <[email protected]>

Looks good,

Reviewed-by: Christoph Hellwig <[email protected]>