2010-02-11 22:04:59

by Ben Myers

[permalink] [raw]
Subject: [PATCH 0/2] commit_metadata export operation v4

Hey Bruce,

Here is the latest version of the knfsd sync changes. I have addressed the
suggestions provided by Christoph, Trond, and Dave. This time rebased atop
for-2.6.34 (I hope). Sorry for the extra noise.

Thanks,
Ben

---

Ben Myers (2):
commit_metadata export operation replacing nfsd_sync_dir
xfs_export_operations.commit_metadata


fs/nfsd/nfs3proc.c | 2 -
fs/nfsd/nfs4proc.c | 2 -
fs/nfsd/nfs4recover.c | 4 -
fs/nfsd/nfs4state.c | 2 -
fs/nfsd/nfsproc.c | 4 +
fs/nfsd/vfs.c | 116 +++++++++++++++++++----------------------
fs/nfsd/vfs.h | 3 -
fs/xfs/linux-2.6/xfs_export.c | 43 +++++++++++++++
include/linux/exportfs.h | 6 ++
9 files changed, 111 insertions(+), 71 deletions(-)

--


2010-02-11 22:05:04

by Ben Myers

[permalink] [raw]
Subject: [PATCH 1/2] commit_metadata export operation replacing nfsd_sync_dir

- Add commit_metadata export_operation to allow the underlying filesystem to
decide how to sync parent and child inodes most efficiently.

- Usage of nfsd_sync_dir and write_inode_now has been replaced with the
commit_metadata function that takes a svc_fh and optional dentry for the child.

- The commit_metadata function calls the commit_metadata export_op if it's
there, or else falls back to sync_inode instead of fsync and write_inode_now
because only metadata need be synched here.

- nfsd4_sync_rec_dir now uses vfs_fsync so that commit_metadata can be static

- Add a 'delay_commit' arg to nfsd_setattr so that callers of
nfsd_create_setattr can commit parent and child together avoiding an extra
sync.

Signed-off-by: Ben Myers <[email protected]>
---
fs/nfsd/nfs3proc.c | 2 -
fs/nfsd/nfs4proc.c | 2 -
fs/nfsd/nfs4recover.c | 4 --
fs/nfsd/nfs4state.c | 2 -
fs/nfsd/nfsproc.c | 4 +-
fs/nfsd/vfs.c | 116 ++++++++++++++++++++++------------------------
fs/nfsd/vfs.h | 3 -
include/linux/exportfs.h | 6 ++
8 files changed, 68 insertions(+), 71 deletions(-)

diff --git a/fs/nfsd/nfs3proc.c b/fs/nfsd/nfs3proc.c
index 3d68f45..fe3af23 100644
--- a/fs/nfsd/nfs3proc.c
+++ b/fs/nfsd/nfs3proc.c
@@ -76,7 +76,7 @@ nfsd3_proc_setattr(struct svc_rqst *rqstp, struct nfsd3_sattrargs *argp,

fh_copy(&resp->fh, &argp->fh);
nfserr = nfsd_setattr(rqstp, &resp->fh, &argp->attrs,
- argp->check_guard, argp->guardtime);
+ argp->check_guard, argp->guardtime, 0);
RETURN_STATUS(nfserr);
}

diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
index 37514c4..c6e3c7f 100644
--- a/fs/nfsd/nfs4proc.c
+++ b/fs/nfsd/nfs4proc.c
@@ -803,7 +803,7 @@ nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
if (status)
goto out;
status = nfsd_setattr(rqstp, &cstate->current_fh, &setattr->sa_iattr,
- 0, (time_t)0);
+ 0, 0, 0);
out:
mnt_drop_write(cstate->current_fh.fh_export->ex_path.mnt);
return status;
diff --git a/fs/nfsd/nfs4recover.c b/fs/nfsd/nfs4recover.c
index 5a754f7..98fb98e 100644
--- a/fs/nfsd/nfs4recover.c
+++ b/fs/nfsd/nfs4recover.c
@@ -119,9 +119,7 @@ out_no_tfm:
static void
nfsd4_sync_rec_dir(void)
{
- mutex_lock(&rec_dir.dentry->d_inode->i_mutex);
- nfsd_sync_dir(rec_dir.dentry);
- mutex_unlock(&rec_dir.dentry->d_inode->i_mutex);
+ vfs_fsync(NULL, rec_dir.dentry, 0);
}

int
diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 3a20c09..f2fc8c8 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -2290,7 +2290,7 @@ nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
return 0;
if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
return nfserr_inval;
- return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
+ return nfsd_setattr(rqstp, fh, &iattr, 0, 0, 0);
}

static __be32
diff --git a/fs/nfsd/nfsproc.c b/fs/nfsd/nfsproc.c
index a047ad6..f5e8280 100644
--- a/fs/nfsd/nfsproc.c
+++ b/fs/nfsd/nfsproc.c
@@ -69,7 +69,7 @@ nfsd_proc_setattr(struct svc_rqst *rqstp, struct nfsd_sattrargs *argp,
argp->attrs.ia_valid, (long) argp->attrs.ia_size);

fh_copy(&resp->fh, &argp->fh);
- nfserr = nfsd_setattr(rqstp, &resp->fh, &argp->attrs,0, (time_t)0);
+ nfserr = nfsd_setattr(rqstp, &resp->fh, &argp->attrs, 0, 0, 0);
return nfsd_return_attrs(nfserr, resp);
}

@@ -326,7 +326,7 @@ nfsd_proc_create(struct svc_rqst *rqstp, struct nfsd_createargs *argp,
*/
attr->ia_valid &= ATTR_SIZE;
if (attr->ia_valid)
- nfserr = nfsd_setattr(rqstp, newfhp, attr, 0, (time_t)0);
+ nfserr = nfsd_setattr(rqstp, newfhp, attr, 0, 0, 0);
}

out_unlock:
diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
index ed024d3..97474fb 100644
--- a/fs/nfsd/vfs.c
+++ b/fs/nfsd/vfs.c
@@ -27,6 +27,8 @@
#include <linux/jhash.h>
#include <linux/ima.h>
#include <asm/uaccess.h>
+#include <linux/exportfs.h>
+#include <linux/writeback.h>

#ifdef CONFIG_NFSD_V3
#include "xdr3.h"
@@ -271,6 +273,38 @@ out:
return err;
}

+/*
+ * Commit metadata changes to stable storage. You pay pass NULL for dchild.
+ */
+static int
+commit_metadata(struct svc_fh *fhp, struct dentry *dchild)
+{
+ struct inode *parent = fhp->fh_dentry->d_inode;
+ struct inode *child = NULL;
+ const struct export_operations *export_ops = parent->i_sb->s_export_op;
+ struct writeback_control wbc = {
+ .sync_mode = WB_SYNC_ALL,
+ .nr_to_write = 0, /* metadata only */
+ };
+ int error = 0, error2 = 0;
+
+ if (!EX_ISSYNC(fhp->fh_export))
+ return 0;
+
+ if (dchild)
+ child = dchild->d_inode;
+
+ if (export_ops->commit_metadata) {
+ error = export_ops->commit_metadata(parent, child);
+ } else {
+ if (child)
+ error2 = sync_inode(child, &wbc);
+ error = sync_inode(parent, &wbc);
+ if (error2)
+ error = error2;
+ }
+ return error;
+}

/*
* Set various file attributes.
@@ -278,7 +312,7 @@ out:
*/
__be32
nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
- int check_guard, time_t guardtime)
+ int check_guard, time_t guardtime, int delay_commit)
{
struct dentry *dentry;
struct inode *inode;
@@ -415,9 +449,8 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
}
if (size_change)
put_write_access(inode);
- if (!err)
- if (EX_ISSYNC(fhp->fh_export))
- write_inode_now(inode, 1);
+ if (!err && !delay_commit)
+ err = nfserrno(commit_metadata(fhp, NULL));
out:
return err;

@@ -769,28 +802,6 @@ nfsd_close(struct file *filp)
}

/*
- * Sync a directory to disk.
- *
- * We can't just call vfs_fsync because our requirements are slightly odd:
- *
- * a) we do not have a file struct available
- * b) we expect to have i_mutex already held by the caller
- */
-int
-nfsd_sync_dir(struct dentry *dentry)
-{
- struct inode *inode = dentry->d_inode;
- int error;
-
- WARN_ON(!mutex_is_locked(&inode->i_mutex));
-
- error = filemap_write_and_wait(inode->i_mapping);
- if (!error && inode->i_fop->fsync)
- error = inode->i_fop->fsync(NULL, dentry, 0);
- return error;
-}
-
-/*
* Obtain the readahead parameters for the file
* specified by (dev, ino).
*/
@@ -1198,8 +1209,10 @@ nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp,
*/
if (current_fsuid() != 0)
iap->ia_valid &= ~(ATTR_UID|ATTR_GID);
- if (iap->ia_valid)
- return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0);
+ if (iap->ia_valid) {
+ return nfsd_setattr(rqstp, resfhp, iap, 0, 0,
+ 1 /* delay commit. our caller does it. */);
+ }
return 0;
}

@@ -1331,12 +1344,8 @@ nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp,
goto out_nfserr;
}

- if (EX_ISSYNC(fhp->fh_export)) {
- err = nfserrno(nfsd_sync_dir(dentry));
- write_inode_now(dchild->d_inode, 1);
- }
-
- err2 = nfsd_create_setattr(rqstp, resfhp, iap);
+ err = nfsd_create_setattr(rqstp, resfhp, iap);
+ err2 = nfserrno(commit_metadata(fhp, dchild));
if (err2)
err = err2;
mnt_drop_write(fhp->fh_export->ex_path.mnt);
@@ -1368,7 +1377,6 @@ nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp,
struct dentry *dentry, *dchild = NULL;
struct inode *dirp;
__be32 err;
- __be32 err2;
int host_err;
__u32 v_mtime=0, v_atime=0;

@@ -1463,11 +1471,6 @@ nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp,
if (created)
*created = 1;

- if (EX_ISSYNC(fhp->fh_export)) {
- err = nfserrno(nfsd_sync_dir(dentry));
- /* setattr will sync the child (or not) */
- }
-
nfsd_check_ignore_resizing(iap);

if (createmode == NFS3_CREATE_EXCLUSIVE) {
@@ -1482,9 +1485,9 @@ nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp,
}

set_attr:
- err2 = nfsd_create_setattr(rqstp, resfhp, iap);
- if (err2)
- err = err2;
+ err = nfsd_create_setattr(rqstp, resfhp, iap);
+ if (!err)
+ err = nfserrno(commit_metadata(fhp, dchild));

mnt_drop_write(fhp->fh_export->ex_path.mnt);
/*
@@ -1599,12 +1602,10 @@ nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp,
}
} else
host_err = vfs_symlink(dentry->d_inode, dnew, path);
-
- if (!host_err) {
- if (EX_ISSYNC(fhp->fh_export))
- host_err = nfsd_sync_dir(dentry);
- }
err = nfserrno(host_err);
+ if (!err)
+ err = nfserrno(commit_metadata(fhp, NULL));
+
fh_unlock(fhp);

mnt_drop_write(fhp->fh_export->ex_path.mnt);
@@ -1666,11 +1667,7 @@ nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp,
}
host_err = vfs_link(dold, dirp, dnew);
if (!host_err) {
- if (EX_ISSYNC(ffhp->fh_export)) {
- err = nfserrno(nfsd_sync_dir(ddir));
- write_inode_now(dest, 1);
- }
- err = 0;
+ err = nfserrno(commit_metadata(ffhp, dold));
} else {
if (host_err == -EXDEV && rqstp->rq_vers == 2)
err = nfserr_acces;
@@ -1766,10 +1763,10 @@ nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
goto out_dput_new;

host_err = vfs_rename(fdir, odentry, tdir, ndentry);
- if (!host_err && EX_ISSYNC(tfhp->fh_export)) {
- host_err = nfsd_sync_dir(tdentry);
+ if (!host_err) {
+ host_err = commit_metadata(tfhp, NULL);
if (!host_err)
- host_err = nfsd_sync_dir(fdentry);
+ host_err = commit_metadata(ffhp, NULL);
}

mnt_drop_write(ffhp->fh_export->ex_path.mnt);
@@ -1850,12 +1847,9 @@ nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,

dput(rdentry);

- if (host_err)
- goto out_drop;
- if (EX_ISSYNC(fhp->fh_export))
- host_err = nfsd_sync_dir(dentry);
+ if (!host_err)
+ host_err = commit_metadata(fhp, NULL);

-out_drop:
mnt_drop_write(fhp->fh_export->ex_path.mnt);
out_nfserr:
err = nfserrno(host_err);
diff --git a/fs/nfsd/vfs.h b/fs/nfsd/vfs.h
index 4b1de0a..5062afd 100644
--- a/fs/nfsd/vfs.h
+++ b/fs/nfsd/vfs.h
@@ -41,7 +41,7 @@ __be32 nfsd_lookup_dentry(struct svc_rqst *, struct svc_fh *,
const char *, unsigned int,
struct svc_export **, struct dentry **);
__be32 nfsd_setattr(struct svc_rqst *, struct svc_fh *,
- struct iattr *, int, time_t);
+ struct iattr *, int, time_t, int);
int nfsd_mountpoint(struct dentry *, struct svc_export *);
#ifdef CONFIG_NFSD_V4
__be32 nfsd4_set_nfs4_acl(struct svc_rqst *, struct svc_fh *,
@@ -91,7 +91,6 @@ __be32 nfsd_statfs(struct svc_rqst *, struct svc_fh *,
int nfsd_notify_change(struct inode *, struct iattr *);
__be32 nfsd_permission(struct svc_rqst *, struct svc_export *,
struct dentry *, int);
-int nfsd_sync_dir(struct dentry *dp);

#if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
struct posix_acl *nfsd_get_posix_acl(struct svc_fh *, int);
diff --git a/include/linux/exportfs.h b/include/linux/exportfs.h
index dc12f41..9102ecf 100644
--- a/include/linux/exportfs.h
+++ b/include/linux/exportfs.h
@@ -96,6 +96,7 @@ struct fid {
* @fh_to_parent: find the implied object's parent and get a dentry for it
* @get_name: find the name for a given inode in a given directory
* @get_parent: find the parent of a given directory
+ * @commit_metadata: commit metadata changes to stable storage
*
* See Documentation/filesystems/nfs/Exporting for details on how to use
* this interface correctly.
@@ -137,6 +138,10 @@ struct fid {
* is also a directory. In the event that it cannot be found, or storage
* space cannot be allocated, a %ERR_PTR should be returned.
*
+ * commit_metadata:
+ * @commit_metadata should commit metadata changes to stable storage.
+ * Parent or child can be NULL.
+ *
* Locking rules:
* get_parent is called with child->d_inode->i_mutex down
* get_name is not (which is possibly inconsistent)
@@ -152,6 +157,7 @@ struct export_operations {
int (*get_name)(struct dentry *parent, char *name,
struct dentry *child);
struct dentry * (*get_parent)(struct dentry *child);
+ int (*commit_metadata)(struct inode *parent, struct inode *child);
};

extern int exportfs_encode_fh(struct dentry *dentry, struct fid *fid,


2010-02-11 22:05:13

by Ben Myers

[permalink] [raw]
Subject: [PATCH 2/2] xfs_export_operations.commit_metadata

This is the commit_metadata export operation for XFS, including the changes
suggested by hch and dgc:

- Takes two two inodes instead of dentries and can assume the parent is
always set.

- Uses xfs_lock_two_inodes for the ilock.

- Forces the log up to the larger lsn of parent and child.

- Uses XFS_LSN_CMP for lsn comparison.

- Doesn't force the log if nobody had a pincount.

Signed-off-by: Ben Myers <[email protected]>
---
fs/xfs/linux-2.6/xfs_export.c | 43 +++++++++++++++++++++++++++++++++++++++++
1 files changed, 43 insertions(+), 0 deletions(-)

diff --git a/fs/xfs/linux-2.6/xfs_export.c b/fs/xfs/linux-2.6/xfs_export.c
index 87b8cbd..d37ced1 100644
--- a/fs/xfs/linux-2.6/xfs_export.c
+++ b/fs/xfs/linux-2.6/xfs_export.c
@@ -29,6 +29,7 @@
#include "xfs_vnodeops.h"
#include "xfs_bmap_btree.h"
#include "xfs_inode.h"
+#include "xfs_inode_item.h"

/*
* Note that we only accept fileids which are long enough rather than allow
@@ -215,9 +216,51 @@ xfs_fs_get_parent(
return d_obtain_alias(VFS_I(cip));
}

+STATIC int
+xfs_fs_nfs_commit_metadata(
+ struct inode *parent,
+ struct inode *child)
+{
+ struct xfs_inode *dp = XFS_I(parent);
+ struct xfs_inode *ip = NULL;
+ struct xfs_mount *mp = dp->i_mount;
+ xfs_lsn_t force_lsn = NULLCOMMITLSN;
+ int error = 0;
+
+ if (child) {
+ ip = XFS_I(child);
+ xfs_lock_two_inodes(dp, ip, XFS_ILOCK_SHARED);
+ } else {
+ xfs_ilock(dp, XFS_ILOCK_SHARED);
+ }
+
+ if (xfs_ipincount(dp)) {
+ force_lsn = dp->i_itemp->ili_last_lsn;
+ if (ip && xfs_ipincount(ip) &&
+ XFS_LSN_CMP(force_lsn, ip->i_itemp->ili_last_lsn) < 0) {
+ force_lsn = ip->i_itemp->ili_last_lsn;
+ }
+ } else if (ip && xfs_ipincount(ip)) {
+ force_lsn = ip->i_itemp->ili_last_lsn;
+ }
+
+ if (force_lsn != NULLCOMMITLSN) {
+ error = _xfs_log_force(mp, force_lsn,
+ XFS_LOG_FORCE | XFS_LOG_SYNC, NULL);
+ }
+
+ if (child)
+ xfs_iunlock(ip, XFS_ILOCK_SHARED);
+ if (parent)
+ xfs_iunlock(dp, XFS_ILOCK_SHARED);
+
+ return error;
+}
+
const struct export_operations xfs_export_operations = {
.encode_fh = xfs_fs_encode_fh,
.fh_to_dentry = xfs_fs_fh_to_dentry,
.fh_to_parent = xfs_fs_fh_to_parent,
.get_parent = xfs_fs_get_parent,
+ .commit_metadata = xfs_fs_nfs_commit_metadata,
};


2010-02-12 14:23:45

by Alex Elder

[permalink] [raw]
Subject: Re: [PATCH 1/2] commit_metadata export operation replacing nfsd_sync_dir

Generally this looks like a good change, but I do have
some questions embedded below.

-Alex

On Thu, 2010-02-11 at 16:05 -0600, Ben Myers wrote:
> - Add commit_metadata export_operation to allow the underlying filesystem to
> decide how to sync parent and child inodes most efficiently.
>
> - Usage of nfsd_sync_dir and write_inode_now has been replaced with the
> commit_metadata function that takes a svc_fh and optional dentry for the child.

In the cases you pass the child dentry, you will now be syncing
things to disk in a slightly different way/order from before.
I think this way is actually much better, but I thought I'd ask
anyway: are you sure that it's OK to sync them at the same time,
rather than directory first, then child (or the other way around,
depending on the case)?

> - The commit_metadata function calls the commit_metadata export_op if it's
> there, or else falls back to sync_inode instead of fsync and write_inode_now
> because only metadata need be synched here.
>
> - nfsd4_sync_rec_dir now uses vfs_fsync so that commit_metadata can be static
>
> - Add a 'delay_commit' arg to nfsd_setattr so that callers of
> nfsd_create_setattr can commit parent and child together avoiding an extra
> sync.

I think this is a distinct enough change to warrant a separate patch
from the commit_metadata feature.

>
> Signed-off-by: Ben Myers <[email protected]>

. . .

> index 5a754f7..98fb98e 100644
> --- a/fs/nfsd/nfs4recover.c
> +++ b/fs/nfsd/nfs4recover.c
> @@ -119,9 +119,7 @@ out_no_tfm:
> static void
> nfsd4_sync_rec_dir(void)
> {
> - mutex_lock(&rec_dir.dentry->d_inode->i_mutex);
> - nfsd_sync_dir(rec_dir.dentry);
> - mutex_unlock(&rec_dir.dentry->d_inode->i_mutex);
> + vfs_fsync(NULL, rec_dir.dentry, 0);

Why do you no longer need to acquire the mutex here?
I see it gets acquired during the ->fsync() call
inside vfs_fsync_range(), but is there a reason
that it needed to be held during the filemap_write_and_wait()
(as is done in the nfsd_sync_dir() code) also?

. . .

> diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
> index ed024d3..97474fb 100644
> --- a/fs/nfsd/vfs.c
> +++ b/fs/nfsd/vfs.c

. . .

> @@ -415,9 +449,8 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
> }
> if (size_change)
> put_write_access(inode);
> - if (!err)
> - if (EX_ISSYNC(fhp->fh_export))
> - write_inode_now(inode, 1);
> + if (!err && !delay_commit)
> + err = nfserrno(commit_metadata(fhp, NULL));

Is this sufficient even if the size has changed?

> out:
> return err;
>

. . .

> @@ -1766,10 +1763,10 @@ nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
> goto out_dput_new;
>
> host_err = vfs_rename(fdir, odentry, tdir, ndentry);
> - if (!host_err && EX_ISSYNC(tfhp->fh_export)) {
> - host_err = nfsd_sync_dir(tdentry);
> + if (!host_err) {
> + host_err = commit_metadata(tfhp, NULL);
> if (!host_err)
> - host_err = nfsd_sync_dir(fdentry);
> + host_err = commit_metadata(ffhp, NULL);

It violates the spirit of the "parent" and "child"
nature of its arguments, but it might be nice to
commit both directories' metadata with the same
call here.

. . .


2010-02-12 14:46:51

by Alex Elder

[permalink] [raw]
Subject: Re: [PATCH 2/2] xfs_export_operations.commit_metadata

On Thu, 2010-02-11 at 16:05 -0600, Ben Myers wrote:
> This is the commit_metadata export operation for XFS, including the changes
> suggested by hch and dgc:
>
> - Takes two two inodes instead of dentries and can assume the parent is
> always set.

I alluded to this in my review of the first patch.
This could be changed considered in a more generic
sense, "sync one or two inodes' metadata" rather
than presupposing the two inodes have a parent/child
relationship.

> - Uses xfs_lock_two_inodes for the ilock.
>
> - Forces the log up to the larger lsn of parent and child.
>
> - Uses XFS_LSN_CMP for lsn comparison.
>
> - Doesn't force the log if nobody had a pincount.

So if the log doesn't get forced, what causes the
desired metadata sync expected as a result of this
call? (Maybe this is a dumb question.)

> Signed-off-by: Ben Myers <[email protected]>
> ---
> fs/xfs/linux-2.6/xfs_export.c | 43 +++++++++++++++++++++++++++++++++++++++++
> 1 files changed, 43 insertions(+), 0 deletions(-)
>
> diff --git a/fs/xfs/linux-2.6/xfs_export.c b/fs/xfs/linux-2.6/xfs_export.c
> index 87b8cbd..d37ced1 100644
> --- a/fs/xfs/linux-2.6/xfs_export.c
> +++ b/fs/xfs/linux-2.6/xfs_export.c
> @@ -29,6 +29,7 @@
> #include "xfs_vnodeops.h"
> #include "xfs_bmap_btree.h"
> #include "xfs_inode.h"
> +#include "xfs_inode_item.h"
>
> /*
> * Note that we only accept fileids which are long enough rather than allow
> @@ -215,9 +216,51 @@ xfs_fs_get_parent(
> return d_obtain_alias(VFS_I(cip));
> }
>
> +STATIC int
> +xfs_fs_nfs_commit_metadata(
> + struct inode *parent,
> + struct inode *child)
> +{
> + struct xfs_inode *dp = XFS_I(parent);
> + struct xfs_inode *ip = NULL;
> + struct xfs_mount *mp = dp->i_mount;
> + xfs_lsn_t force_lsn = NULLCOMMITLSN;
> + int error = 0;
> +
> + if (child) {
> + ip = XFS_I(child);
> + xfs_lock_two_inodes(dp, ip, XFS_ILOCK_SHARED);
> + } else {
> + xfs_ilock(dp, XFS_ILOCK_SHARED);
> + }
> +
> + if (xfs_ipincount(dp)) {
> + force_lsn = dp->i_itemp->ili_last_lsn;
> + if (ip && xfs_ipincount(ip) &&
> + XFS_LSN_CMP(force_lsn, ip->i_itemp->ili_last_lsn) < 0) {
> + force_lsn = ip->i_itemp->ili_last_lsn;
> + }
> + } else if (ip && xfs_ipincount(ip)) {
> + force_lsn = ip->i_itemp->ili_last_lsn;
> + }
> +
> + if (force_lsn != NULLCOMMITLSN) {
> + error = _xfs_log_force(mp, force_lsn,
> + XFS_LOG_FORCE | XFS_LOG_SYNC, NULL);

You want this here:
error = xfs_log_force_lsn(mp, force_lsn, XFS_LOG_FORCE | XFS_LOG_SYNC);


> + }
> +
> + if (child)
> + xfs_iunlock(ip, XFS_ILOCK_SHARED);
> + if (parent)
> + xfs_iunlock(dp, XFS_ILOCK_SHARED);
> +
> + return error;
> +}
> +
> const struct export_operations xfs_export_operations = {
> .encode_fh = xfs_fs_encode_fh,
> .fh_to_dentry = xfs_fs_fh_to_dentry,
> .fh_to_parent = xfs_fs_fh_to_parent,
> .get_parent = xfs_fs_get_parent,
> + .commit_metadata = xfs_fs_nfs_commit_metadata,
> };
>
> _______________________________________________
> xfs mailing list
> [email protected]
> http://oss.sgi.com/mailman/listinfo/xfs




2010-02-12 17:31:09

by Christoph Hellwig

[permalink] [raw]
Subject: Re: [PATCH 1/2] commit_metadata export operation replacing nfsd_sync_dir

> In the cases you pass the child dentry, you will now be syncing
> things to disk in a slightly different way/order from before.
> I think this way is actually much better, but I thought I'd ask
> anyway: are you sure that it's OK to sync them at the same time,
> rather than directory first, then child (or the other way around,
> depending on the case)?

The order is defined by when we commit the transactions, if we do
the log force on the later one first we already write out the first
transaction. Note that in any transaction filesystems the changes
that create/mkdir/link/unlink do to parent and child will be in
the same transaction anyway, which is kinda the point of adding
the transactions to start with. Only for the case where we do
a setattr in addition to the primary transaction we'll actually
have another transaction to deal with and the above applies.

> > nfsd4_sync_rec_dir(void)
> > {
> > - mutex_lock(&rec_dir.dentry->d_inode->i_mutex);
> > - nfsd_sync_dir(rec_dir.dentry);
> > - mutex_unlock(&rec_dir.dentry->d_inode->i_mutex);
> > + vfs_fsync(NULL, rec_dir.dentry, 0);
>
> Why do you no longer need to acquire the mutex here?
> I see it gets acquired during the ->fsync() call
> inside vfs_fsync_range(), but is there a reason
> that it needed to be held during the filemap_write_and_wait()
> (as is done in the nfsd_sync_dir() code) also?

We don't need i_mutex for filemap_write_and_wait, it's just held
because someone used the nfsd_sync_dir helper where it doesn't
fit very well.

> > @@ -415,9 +449,8 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
> > }
> > if (size_change)
> > put_write_access(inode);
> > - if (!err)
> > - if (EX_ISSYNC(fhp->fh_export))
> > - write_inode_now(inode, 1);
> > + if (!err && !delay_commit)
> > + err = nfserrno(commit_metadata(fhp, NULL));
>
> Is this sufficient even if the size has changed?

Yes.


2010-02-12 17:47:10

by Christoph Hellwig

[permalink] [raw]
Subject: Re: [PATCH 2/2] xfs_export_operations.commit_metadata

On Fri, Feb 12, 2010 at 08:46:46AM -0600, Alex Elder wrote:
> On Thu, 2010-02-11 at 16:05 -0600, Ben Myers wrote:
> > This is the commit_metadata export operation for XFS, including the changes
> > suggested by hch and dgc:
> >
> > - Takes two two inodes instead of dentries and can assume the parent is
> > always set.
>
> I alluded to this in my review of the first patch.
> This could be changed considered in a more generic
> sense, "sync one or two inodes' metadata" rather
> than presupposing the two inodes have a parent/child
> relationship.

Or just implement my later suggestion to only pass one inode to the
method and cal in on both inodes. For the non-create case where
we have only one transaction to deal with fhr first call will take
care of it and unpin the second inode by forcing the log buffer out.
For the create case we need to make sure to call it on the child first
so that we force out the setattr transaction which also forced out the
earlier one. This keeps the calling convention quite a bit simpler,
and also means we don't have to bother with locking two inodes or lsn
comparisms.

> > - Uses xfs_lock_two_inodes for the ilock.
> >
> > - Forces the log up to the larger lsn of parent and child.
> >
> > - Uses XFS_LSN_CMP for lsn comparison.
> >
> > - Doesn't force the log if nobody had a pincount.
>
> So if the log doesn't get forced, what causes the
> desired metadata sync expected as a result of this
> call? (Maybe this is a dumb question.)

Inodes get pinned on transaction commit, and unpinned when the log I/O
for that transaction completes. If the inode is not pinned this implies
it has already been written to disk, e.g. because we're filling the log
so fast that we need to write out more log buffers in that tiny window
between the metada operation and the commit_metadata call.

> > + force_lsn = ip->i_itemp->ili_last_lsn;
> > + }
> > +
> > + if (force_lsn != NULLCOMMITLSN) {
> > + error = _xfs_log_force(mp, force_lsn,
> > + XFS_LOG_FORCE | XFS_LOG_SYNC, NULL);
>
> You want this here:
> error = xfs_log_force_lsn(mp, force_lsn, XFS_LOG_FORCE | XFS_LOG_SYNC);

In the XFS tree we do want either

xfs_log_force_lsn(mp, force_lsn, XFS_LOG_SYNC);

or
error = _xfs_log_force_lsn(mp, force_lsn, XFS_LOG_SYNC, NULL);

if we care enough about the returned error. But Ben is working against
the NFS tree which doesn't have that change yet.

We can deal with that by either commiting the old variant to the nfs
tree and then leaving sending Stephen a patch to fix it up in -next,
or just not apply the xfs commit_metadata implementation yet, and wait
for it until both the xfs and nfs trees have hit mainline.


2010-02-12 19:56:11

by Ben Myers

[permalink] [raw]
Subject: Re: [PATCH 2/2] xfs_export_operations.commit_metadata

On Fri, Feb 12, 2010 at 12:47:07PM -0500, Christoph Hellwig wrote:
> On Fri, Feb 12, 2010 at 08:46:46AM -0600, Alex Elder wrote:
> > On Thu, 2010-02-11 at 16:05 -0600, Ben Myers wrote:
> > > This is the commit_metadata export operation for XFS, including the changes
> > > suggested by hch and dgc:
> > >
> > > - Takes two two inodes instead of dentries and can assume the parent is
> > > always set.
> >
> > I alluded to this in my review of the first patch.
> > This could be changed considered in a more generic
> > sense, "sync one or two inodes' metadata" rather
> > than presupposing the two inodes have a parent/child
> > relationship.
>
> Or just implement my later suggestion to only pass one inode to the
> method and cal in on both inodes. For the non-create case where
> we have only one transaction to deal with fhr first call will take
> care of it and unpin the second inode by forcing the log buffer out.
> For the create case we need to make sure to call it on the child first
> so that we force out the setattr transaction which also forced out the
> earlier one.

I chose not implement that suggestion because I prefer not to rely upon
the coincidence that the child be modified last and synced first in
knfsd. It is better that the intent of the patch be clear, and that
filesystems other than xfs also have enough information to sort out how
to sync more efficiently. knfsd shouldn't be forced to sync in a
specific order just because that's what works best for xfs. Or, mebbe
it should and I'm being thick. ;)

Alex's suggestion that ->commit_metadata be more generic about the
relationship of the two inodes seems reasonable, not sure what that
means for vfs.c:commit_metadata. This will help with the rename case.

> This keeps the calling convention quite a bit simpler,
> and also means we don't have to bother with locking two inodes or lsn
> comparisms.

Don't need the ilock to check pincount?

> > > + error = _xfs_log_force(mp, force_lsn,
> > > + XFS_LOG_FORCE | XFS_LOG_SYNC, NULL);
> >
> > You want this here:
> > error = xfs_log_force_lsn(mp, force_lsn, XFS_LOG_FORCE | XFS_LOG_SYNC);
>
> In the XFS tree we do want either
>
> xfs_log_force_lsn(mp, force_lsn, XFS_LOG_SYNC);
>
> or
> error = _xfs_log_force_lsn(mp, force_lsn, XFS_LOG_SYNC, NULL);
>
> if we care enough about the returned error. But Ben is working against
> the NFS tree which doesn't have that change yet.
>
> We can deal with that by either commiting the old variant to the nfs
> tree and then leaving sending Stephen a patch to fix it up in -next,
> or just not apply the xfs commit_metadata implementation yet, and wait
> for it until both the xfs and nfs trees have hit mainline.

Yeah. I don't know who Stephen is.

Thanks,
Ben

2010-02-12 20:03:29

by J. Bruce Fields

[permalink] [raw]
Subject: Re: [PATCH 2/2] xfs_export_operations.commit_metadata

On Fri, Feb 12, 2010 at 01:56:47PM -0600, [email protected] wrote:
> On Fri, Feb 12, 2010 at 12:47:07PM -0500, Christoph Hellwig wrote:
> > if we care enough about the returned error. But Ben is working against
> > the NFS tree which doesn't have that change yet.
> >
> > We can deal with that by either commiting the old variant to the nfs
> > tree and then leaving sending Stephen a patch to fix it up in -next,
> > or just not apply the xfs commit_metadata implementation yet, and wait
> > for it until both the xfs and nfs trees have hit mainline.

Assuming the xfs tree's something stable, I could also just go ahead and
pull it into the nfsd tree.

(Or we could do it the other way around--as long as when the merge
window opens we try to send the pull requests in the same order.)

--b.

2010-02-12 20:36:00

by Christoph Hellwig

[permalink] [raw]
Subject: Re: [PATCH 2/2] xfs_export_operations.commit_metadata

On Fri, Feb 12, 2010 at 01:56:47PM -0600, [email protected] wrote:
> I chose not implement that suggestion because I prefer not to rely upon
> the coincidence that the child be modified last and synced first in
> knfsd.

It does ot rely on that coincidence for correctness, just for a small
performance optimization.

> It is better that the intent of the patch be clear, and that
> filesystems other than xfs also have enough information to sort out how
> to sync more efficiently. knfsd shouldn't be forced to sync in a
> specific order just because that's what works best for xfs. Or, mebbe
> it should and I'm being thick. ;)

The order of parent and child only happens in the create case where
NFS does a separate ->setattr call. For all the operations handled
by the filesystem parent and child are in the same transaction for
every transactional filesystem.

> > This keeps the calling convention quite a bit simpler,
> > and also means we don't have to bother with locking two inodes or lsn
> > comparisms.
>
> Don't need the ilock to check pincount?

Ah sorry, that should have read not bothering with locking two inodes
at the same time, which always is a bit troublesome. We do need to lock
the inode for looking at the pincount.

> > We can deal with that by either commiting the old variant to the nfs
> > tree and then leaving sending Stephen a patch to fix it up in -next,
> > or just not apply the xfs commit_metadata implementation yet, and wait
> > for it until both the xfs and nfs trees have hit mainline.
>
> Yeah. I don't know who Stephen is.

Stephen Rothwell is the maintainer of the linux-next tree.


2010-02-12 20:37:02

by Christoph Hellwig

[permalink] [raw]
Subject: Re: [PATCH 2/2] xfs_export_operations.commit_metadata

On Fri, Feb 12, 2010 at 03:03:59PM -0500, J. Bruce Fields wrote:
> Assuming the xfs tree's something stable, I could also just go ahead and
> pull it into the nfsd tree.
>
> (Or we could do it the other way around--as long as when the merge
> window opens we try to send the pull requests in the same order.)

The xfs tree is supposed to be stable and we only had very few incidents
where a patch had to be pulled again. But it's also a quite large set
of changes.