From b31268ac793fd300da66b9c28bbf0a200339ab96 Mon Sep 17 00:00:00 2001
From: Trond Myklebust <Trond.Myklebust@netapp.com>
Date: Mon, 21 Mar 2011 17:02:00 -0400
Subject: FS: Use stable writes when not doing a bulk flush
If we're only doing a single write, and there are no other unstable
writes being queued up, we might want to just flip to using a stable
write RPC call.
Reviewed-by: NeilBrown <neilb@suse.de>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/pagelist.c | 3 +++
fs/nfs/write.c | 16 +++++++++++++---
2 files changed, 16 insertions(+), 3 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c
index 23e794410669..fd85618149a1 100644
--- a/fs/nfs/pagelist.c
+++ b/fs/nfs/pagelist.c
@@ -223,6 +223,7 @@ void nfs_pageio_init(struct nfs_pageio_descriptor *desc,
desc->pg_count = 0;
desc->pg_bsize = bsize;
desc->pg_base = 0;
+ desc->pg_moreio = 0;
desc->pg_inode = inode;
desc->pg_doio = doio;
desc->pg_ioflags = io_flags;
@@ -335,9 +336,11 @@ int nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
struct nfs_page *req)
{
while (!nfs_pageio_do_add_request(desc, req)) {
+ desc->pg_moreio = 1;
nfs_pageio_doio(desc);
if (desc->pg_error < 0)
return 0;
+ desc->pg_moreio = 0;
}
return 1;
}
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 47a3ad63e0d5..4d686ee53244 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -179,8 +179,8 @@ static int wb_priority(struct writeback_control *wbc)
if (wbc->for_reclaim)
return FLUSH_HIGHPRI | FLUSH_STABLE;
if (wbc->for_kupdate || wbc->for_background)
- return FLUSH_LOWPRI;
- return 0;
+ return FLUSH_LOWPRI | FLUSH_COND_STABLE;
+ return FLUSH_COND_STABLE;
}
/*
@@ -863,7 +863,7 @@ static int nfs_write_rpcsetup(struct nfs_page *req,
data->args.context = get_nfs_open_context(req->wb_context);
data->args.lock_context = req->wb_lock_context;
data->args.stable = NFS_UNSTABLE;
- if (how & FLUSH_STABLE) {
+ if (how & (FLUSH_STABLE | FLUSH_COND_STABLE)) {
data->args.stable = NFS_DATA_SYNC;
if (!nfs_need_commit(NFS_I(inode)))
data->args.stable = NFS_FILE_SYNC;
@@ -912,6 +912,12 @@ static int nfs_flush_multi(struct nfs_pageio_descriptor *desc)
nfs_list_remove_request(req);
+ if ((desc->pg_ioflags & FLUSH_COND_STABLE) &&
+ (desc->pg_moreio || NFS_I(desc->pg_inode)->ncommit ||
+ desc->pg_count > wsize))
+ desc->pg_ioflags &= ~FLUSH_COND_STABLE;
+
+
nbytes = desc->pg_count;
do {
size_t len = min(nbytes, wsize);
@@ -1002,6 +1008,10 @@ static int nfs_flush_one(struct nfs_pageio_descriptor *desc)
if ((!lseg) && list_is_singular(&data->pages))
lseg = pnfs_update_layout(desc->pg_inode, req->wb_context, IOMODE_RW);
+ if ((desc->pg_ioflags & FLUSH_COND_STABLE) &&
+ (desc->pg_moreio || NFS_I(desc->pg_inode)->ncommit))
+ desc->pg_ioflags &= ~FLUSH_COND_STABLE;
+
/* Set up the argument struct */
ret = nfs_write_rpcsetup(req, data, &nfs_write_full_ops, desc->pg_count, 0, lseg, desc->pg_ioflags);
out:
--
cgit v1.2.3
From b8413f98f997bb3ed7327e6d7117e7e91ce010c3 Mon Sep 17 00:00:00 2001
From: Trond Myklebust <Trond.Myklebust@netapp.com>
Date: Mon, 21 Mar 2011 15:37:01 -0400
Subject: NFS: Fix a hang/infinite loop in nfs_wb_page()
When one of the two waits in nfs_commit_inode() is interrupted, it
returns a non-negative value, which causes nfs_wb_page() to think
that the operation was successful causing it to busy-loop rather
than exiting.
It also causes nfs_file_fsync() to incorrectly report the file as
being successfully committed to disk.
This patch fixes both problems by ensuring that we return an error
if the attempts to wait fail.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: stable@kernel.org
---
fs/nfs/write.c | 31 +++++++++++++++++++------------
1 file changed, 19 insertions(+), 12 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 4d686ee53244..55a8c3671233 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -1261,13 +1261,17 @@ void nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
static int nfs_commit_set_lock(struct nfs_inode *nfsi, int may_wait)
{
+ int ret;
+
if (!test_and_set_bit(NFS_INO_COMMIT, &nfsi->flags))
return 1;
- if (may_wait && !out_of_line_wait_on_bit_lock(&nfsi->flags,
- NFS_INO_COMMIT, nfs_wait_bit_killable,
- TASK_KILLABLE))
- return 1;
- return 0;
+ if (!may_wait)
+ return 0;
+ ret = out_of_line_wait_on_bit_lock(&nfsi->flags,
+ NFS_INO_COMMIT,
+ nfs_wait_bit_killable,
+ TASK_KILLABLE);
+ return (ret < 0) ? ret : 1;
}
static void nfs_commit_clear_lock(struct nfs_inode *nfsi)
@@ -1443,9 +1447,10 @@ int nfs_commit_inode(struct inode *inode, int how)
{
LIST_HEAD(head);
int may_wait = how & FLUSH_SYNC;
- int res = 0;
+ int res;
- if (!nfs_commit_set_lock(NFS_I(inode), may_wait))
+ res = nfs_commit_set_lock(NFS_I(inode), may_wait);
+ if (res <= 0)
goto out_mark_dirty;
spin_lock(&inode->i_lock);
res = nfs_scan_commit(inode, &head, 0, 0);
@@ -1454,12 +1459,14 @@ int nfs_commit_inode(struct inode *inode, int how)
int error = nfs_commit_list(inode, &head, how);
if (error < 0)
return error;
- if (may_wait)
- wait_on_bit(&NFS_I(inode)->flags, NFS_INO_COMMIT,
- nfs_wait_bit_killable,
- TASK_KILLABLE);
- else
+ if (!may_wait)
goto out_mark_dirty;
+ error = wait_on_bit(&NFS_I(inode)->flags,
+ NFS_INO_COMMIT,
+ nfs_wait_bit_killable,
+ TASK_KILLABLE);
+ if (error < 0)
+ return error;
} else
nfs_commit_clear_lock(NFS_I(inode));
return res;
--
cgit v1.2.3
From 4667058b772857069378aa8f417203156e315e82 Mon Sep 17 00:00:00 2001
From: Gusev Vitaliy <gusev.vitaliy@nexenta.com>
Date: Wed, 23 Mar 2011 00:40:25 +0300
Subject: nfs4: Fix NULL dereference at d_alloc_and_lookup()
d_alloc_and_lookup() calls i_op->lookup method due to
rootfh changes his fsid.
During mount i_op of NFS root inode is set to
nfs_mountpoint_inode_operations, if rpc_ops->getroot()
and rpc_ops->getattr() return different fsid.
After that nfs_follow_remote_path() raised oops:
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: [< (null)>] (null)
stack trace:
d_alloc_and_lookup+0x4c/0x74
do_lookup+0x1e3/0x280
link_path_walk+0x12e/0xab0
nfs4_remote_get_sb+0x56/0x2c0 [nfs]
path_walk+0x67/0xe0
vfs_path_lookup+0x8e/0x100
nfs_follow_remote_path+0x16f/0x3e0 [nfs]
nfs4_try_mount+0x6f/0xd0 [nfs]
nfs_get_sb+0x269/0x400 [nfs]
vfs_kern_mount+0x8a/0x1f0
do_kern_mount+0x52/0x130
do_mount+0x20a/0x260
sys_mount+0x90/0xe0
system_call_fastpath+0x16/0x1b
So just refresh fsid, as RFC3530 doesn't specify behavior
in case of rootfh changes fsid.
Signed-off-by: Vitaliy Gusev <gusev.vitaliy@nexenta.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/getroot.c | 4 ++++
1 file changed, 4 insertions(+)
(limited to 'fs')
diff --git a/fs/nfs/getroot.c b/fs/nfs/getroot.c
index b5ffe8fa291f..79796521cfdc 100644
--- a/fs/nfs/getroot.c
+++ b/fs/nfs/getroot.c
@@ -199,6 +199,10 @@ struct dentry *nfs4_get_root(struct super_block *sb, struct nfs_fh *mntfh)
goto out;
}
+ if (fattr->valid & NFS_ATTR_FATTR_FSID &&
+ !nfs_fsid_equal(&server->fsid, &fattr->fsid))
+ memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
+
inode = nfs_fhget(sb, mntfh, fattr);
if (IS_ERR(inode)) {
dprintk("nfs_get_root: get root inode failed\n");
--
cgit v1.2.3
From 51e816564db151f7f08c739ad69b9d1dbd166663 Mon Sep 17 00:00:00 2001
From: Jesper Juhl <jj@chaosbits.net>
Date: Mon, 21 Mar 2011 20:35:49 +0100
Subject: Remove pointless memset in nfsacl_encode()
Remove pointless memset() in nfsacl_encode().
Thanks to Trond Myklebust <Trond.Myklebust@netapp.com> for pointing out
that it is not needed since posix_acl_init() will set everything
regardless..
Signed-off-by: Jesper Juhl <jj@chaosbits.net>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs_common/nfsacl.c | 1 -
1 file changed, 1 deletion(-)
(limited to 'fs')
diff --git a/fs/nfs_common/nfsacl.c b/fs/nfs_common/nfsacl.c
index 84c27d69d421..ec0f277be7f5 100644
--- a/fs/nfs_common/nfsacl.c
+++ b/fs/nfs_common/nfsacl.c
@@ -117,7 +117,6 @@ int nfsacl_encode(struct xdr_buf *buf, unsigned int base, struct inode *inode,
* invoked in contexts where a memory allocation failure is
* fatal. Fortunately this fake ACL is small enough to
* construct on the stack. */
- memset(acl2, 0, sizeof(acl2));
posix_acl_init(acl2, 4);
/* Insert entries in canonical order: other orders seem
--
cgit v1.2.3
From e47c085afb3d16cbc6a4bfb10a3b074bb7c58998 Mon Sep 17 00:00:00 2001
From: Trond Myklebust <Trond.Myklebust@netapp.com>
Date: Wed, 23 Mar 2011 08:43:09 -0400
Subject: NFS: Ensure that we update the readdir filp->f_pos correctly
If we're doing a search by readdir cookie, we need to ensure that the
resulting f_pos is updated. To do so, we need to update the
desc->current_index, in the same way that we do in the search by
file offset case.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/dir.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
(limited to 'fs')
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index 2c3eb33b904d..c5c71cb62fbd 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -290,7 +290,6 @@ int nfs_readdir_search_for_pos(struct nfs_cache_array *array, nfs_readdir_descri
if (diff >= array->size) {
if (array->eof_index >= 0)
goto out_eof;
- desc->current_index += array->size;
return -EAGAIN;
}
@@ -311,6 +310,7 @@ int nfs_readdir_search_for_cookie(struct nfs_cache_array *array, nfs_readdir_des
for (i = 0; i < array->size; i++) {
if (array->array[i].cookie == *desc->dir_cookie) {
+ desc->file->f_pos = desc->current_index + i;
desc->cache_entry_index = i;
return 0;
}
@@ -342,6 +342,7 @@ int nfs_readdir_search_array(nfs_readdir_descriptor_t *desc)
if (status == -EAGAIN) {
desc->last_cookie = array->last_cookie;
+ desc->current_index += array->size;
desc->page_index++;
}
nfs_readdir_release_array(desc->page);
--
cgit v1.2.3
From 480c2006ebb44ae03165695db7b3e38c04e0d102 Mon Sep 17 00:00:00 2001
From: Bryan Schumaker <bjschuma@netapp.com>
Date: Wed, 23 Mar 2011 14:48:29 -0400
Subject: NFS: Create nfs_open_dir_context
nfs_opendir() created a context that held much more information than we
need for a readdir. This patch introduces a slimmed-down
nfs_open_dir_context that contains only the cookie and the cred used for
RPC operations. The new context will eventually be used to help detect
readdir loops.
Signed-off-by: Bryan Schumaker <bjschuma@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/dir.c | 54 +++++++++++++++++++++++++++++++++++++++++++-------
fs/nfs/inode.c | 1 -
include/linux/nfs_fs.h | 3 +++
3 files changed, 50 insertions(+), 8 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index c5c71cb62fbd..cda73814f666 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -44,6 +44,7 @@
/* #define NFS_DEBUG_VERBOSE 1 */
static int nfs_opendir(struct inode *, struct file *);
+static int nfs_closedir(struct inode *, struct file *);
static int nfs_readdir(struct file *, void *, filldir_t);
static struct dentry *nfs_lookup(struct inode *, struct dentry *, struct nameidata *);
static int nfs_create(struct inode *, struct dentry *, int, struct nameidata *);
@@ -64,7 +65,7 @@ const struct file_operations nfs_dir_operations = {
.read = generic_read_dir,
.readdir = nfs_readdir,
.open = nfs_opendir,
- .release = nfs_release,
+ .release = nfs_closedir,
.fsync = nfs_fsync_dir,
};
@@ -133,13 +134,33 @@ const struct inode_operations nfs4_dir_inode_operations = {
#endif /* CONFIG_NFS_V4 */
+static struct nfs_open_dir_context *alloc_nfs_open_dir_context(struct rpc_cred *cred)
+{
+ struct nfs_open_dir_context *ctx;
+ ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
+ if (ctx != NULL) {
+ ctx->dir_cookie = 0;
+ ctx->cred = get_rpccred(cred);
+ } else
+ ctx = ERR_PTR(-ENOMEM);
+ return ctx;
+}
+
+static void put_nfs_open_dir_context(struct nfs_open_dir_context *ctx)
+{
+ put_rpccred(ctx->cred);
+ kfree(ctx);
+}
+
/*
* Open file
*/
static int
nfs_opendir(struct inode *inode, struct file *filp)
{
- int res;
+ int res = 0;
+ struct nfs_open_dir_context *ctx;
+ struct rpc_cred *cred;
dfprintk(FILE, "NFS: open dir(%s/%s)\n",
filp->f_path.dentry->d_parent->d_name.name,
@@ -147,8 +168,15 @@ nfs_opendir(struct inode *inode, struct file *filp)
nfs_inc_stats(inode, NFSIOS_VFSOPEN);
- /* Call generic open code in order to cache credentials */
- res = nfs_open(inode, filp);
+ cred = rpc_lookup_cred();
+ if (IS_ERR(cred))
+ return PTR_ERR(cred);
+ ctx = alloc_nfs_open_dir_context(cred);
+ if (IS_ERR(ctx)) {
+ res = PTR_ERR(ctx);
+ goto out;
+ }
+ filp->private_data = ctx;
if (filp->f_path.dentry == filp->f_path.mnt->mnt_root) {
/* This is a mountpoint, so d_revalidate will never
* have been called, so we need to refresh the
@@ -156,9 +184,18 @@ nfs_opendir(struct inode *inode, struct file *filp)
*/
__nfs_revalidate_inode(NFS_SERVER(inode), inode);
}
+out:
+ put_rpccred(cred);
return res;
}
+static int
+nfs_closedir(struct inode *inode, struct file *filp)
+{
+ put_nfs_open_dir_context(filp->private_data);
+ return 0;
+}
+
struct nfs_cache_array_entry {
u64 cookie;
u64 ino;
@@ -355,7 +392,8 @@ static
int nfs_readdir_xdr_filler(struct page **pages, nfs_readdir_descriptor_t *desc,
struct nfs_entry *entry, struct file *file, struct inode *inode)
{
- struct rpc_cred *cred = nfs_file_cred(file);
+ struct nfs_open_dir_context *ctx = file->private_data;
+ struct rpc_cred *cred = ctx->cred;
unsigned long timestamp, gencount;
int error;
@@ -786,6 +824,7 @@ static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
struct inode *inode = dentry->d_inode;
nfs_readdir_descriptor_t my_desc,
*desc = &my_desc;
+ struct nfs_open_dir_context *dir_ctx = filp->private_data;
int res;
dfprintk(FILE, "NFS: readdir(%s/%s) starting at cookie %llu\n",
@@ -802,7 +841,7 @@ static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
memset(desc, 0, sizeof(*desc));
desc->file = filp;
- desc->dir_cookie = &nfs_file_open_context(filp)->dir_cookie;
+ desc->dir_cookie = &dir_ctx->dir_cookie;
desc->decode = NFS_PROTO(inode)->decode_dirent;
desc->plus = NFS_USE_READDIRPLUS(inode);
@@ -854,6 +893,7 @@ static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int origin)
{
struct dentry *dentry = filp->f_path.dentry;
struct inode *inode = dentry->d_inode;
+ struct nfs_open_dir_context *dir_ctx = filp->private_data;
dfprintk(FILE, "NFS: llseek dir(%s/%s, %lld, %d)\n",
dentry->d_parent->d_name.name,
@@ -873,7 +913,7 @@ static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int origin)
}
if (offset != filp->f_pos) {
filp->f_pos = offset;
- nfs_file_open_context(filp)->dir_cookie = 0;
+ dir_ctx->dir_cookie = 0;
}
out:
mutex_unlock(&inode->i_mutex);
diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index 2f8e61816d75..477a2e512b39 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -639,7 +639,6 @@ struct nfs_open_context *alloc_nfs_open_context(struct path *path, struct rpc_cr
ctx->mode = f_mode;
ctx->flags = 0;
ctx->error = 0;
- ctx->dir_cookie = 0;
nfs_init_lock_context(&ctx->lock_context);
ctx->lock_context.open_context = ctx;
INIT_LIST_HEAD(&ctx->list);
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index cb2add401f25..4b851a02f7f3 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -95,7 +95,10 @@ struct nfs_open_context {
int error;
struct list_head list;
+};
+struct nfs_open_dir_context {
+ struct rpc_cred *cred;
__u64 dir_cookie;
};
--
cgit v1.2.3
From 8ef2ce3e16d9bec6cf015207c1c82a5b864046ac Mon Sep 17 00:00:00 2001
From: Bryan Schumaker <bjschuma@netapp.com>
Date: Wed, 23 Mar 2011 15:04:31 -0400
Subject: NFS: Detect loops in a readdir due to bad cookies
Some filesystems (such as ext4) can return the same cookie value for
multiple files. If we try to start a readdir with one of these cookies,
the server will return the first file found with a cookie of the same
value. This can cause the client to enter an infinite loop.
Signed-off-by: Bryan Schumaker <bjschuma@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/dir.c | 28 +++++++++++++++++++++++++++-
include/linux/nfs_fs.h | 2 ++
2 files changed, 29 insertions(+), 1 deletion(-)
(limited to 'fs')
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index cda73814f666..db87a7d1109b 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -139,7 +139,9 @@ static struct nfs_open_dir_context *alloc_nfs_open_dir_context(struct rpc_cred *
struct nfs_open_dir_context *ctx;
ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
if (ctx != NULL) {
+ ctx->duped = 0;
ctx->dir_cookie = 0;
+ ctx->dup_cookie = 0;
ctx->cred = get_rpccred(cred);
} else
ctx = ERR_PTR(-ENOMEM);
@@ -321,6 +323,7 @@ int nfs_readdir_search_for_pos(struct nfs_cache_array *array, nfs_readdir_descri
{
loff_t diff = desc->file->f_pos - desc->current_index;
unsigned int index;
+ struct nfs_open_dir_context *ctx = desc->file->private_data;
if (diff < 0)
goto out_eof;
@@ -333,6 +336,7 @@ int nfs_readdir_search_for_pos(struct nfs_cache_array *array, nfs_readdir_descri
index = (unsigned int)diff;
*desc->dir_cookie = array->array[index].cookie;
desc->cache_entry_index = index;
+ ctx->duped = 0;
return 0;
out_eof:
desc->eof = 1;
@@ -343,11 +347,18 @@ static
int nfs_readdir_search_for_cookie(struct nfs_cache_array *array, nfs_readdir_descriptor_t *desc)
{
int i;
+ loff_t new_pos;
int status = -EAGAIN;
+ struct nfs_open_dir_context *ctx = desc->file->private_data;
for (i = 0; i < array->size; i++) {
if (array->array[i].cookie == *desc->dir_cookie) {
- desc->file->f_pos = desc->current_index + i;
+ new_pos = desc->current_index + i;
+ if (new_pos < desc->file->f_pos) {
+ ctx->dup_cookie = *desc->dir_cookie;
+ ctx->duped = 1;
+ }
+ desc->file->f_pos = new_pos;
desc->cache_entry_index = i;
return 0;
}
@@ -732,6 +743,20 @@ int nfs_do_filldir(nfs_readdir_descriptor_t *desc, void *dirent,
int i = 0;
int res = 0;
struct nfs_cache_array *array = NULL;
+ struct nfs_open_dir_context *ctx = file->private_data;
+
+ if (ctx->duped != 0 && ctx->dup_cookie == *desc->dir_cookie) {
+ if (printk_ratelimit()) {
+ pr_notice("NFS: directory %s/%s contains a readdir loop. "
+ "Please contact your server vendor. "
+ "Offending cookie: %llu\n",
+ file->f_dentry->d_parent->d_name.name,
+ file->f_dentry->d_name.name,
+ *desc->dir_cookie);
+ }
+ res = -ELOOP;
+ goto out;
+ }
array = nfs_readdir_get_array(desc->page);
if (IS_ERR(array)) {
@@ -914,6 +939,7 @@ static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int origin)
if (offset != filp->f_pos) {
filp->f_pos = offset;
dir_ctx->dir_cookie = 0;
+ dir_ctx->duped = 0;
}
out:
mutex_unlock(&inode->i_mutex);
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index 4b851a02f7f3..4179c368844b 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -100,6 +100,8 @@ struct nfs_open_context {
struct nfs_open_dir_context {
struct rpc_cred *cred;
__u64 dir_cookie;
+ __u64 dup_cookie;
+ int duped;
};
/*
--
cgit v1.2.3
From 465d52437d5ce8d4eb9da0d3e3818b51cff163a1 Mon Sep 17 00:00:00 2001
From: Fred Isaman <iisaman@netapp.com>
Date: Wed, 23 Mar 2011 13:27:44 +0000
Subject: NFSv4.1: don't send COMMIT to ds for data sync writes
Based on consensus reached in Feb 2011 interim IETF meeting regarding
use of LAYOUTCOMMIT, it has been decided that a NFS_DATA_SYNC return
from a WRITE to data server should not initiate a COMMIT.
Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/write.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
(limited to 'fs')
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 55a8c3671233..92b4a6614fd3 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -474,7 +474,10 @@ nfs_clear_request_commit(struct nfs_page *req)
static inline
int nfs_write_need_commit(struct nfs_write_data *data)
{
- return data->verf.committed != NFS_FILE_SYNC;
+ if (data->verf.committed == NFS_DATA_SYNC)
+ return data->lseg == NULL;
+ else
+ return data->verf.committed != NFS_FILE_SYNC;
}
static inline
--
cgit v1.2.3
From 9ace33cdc6562699a780b4716d9d52c55dd2151a Mon Sep 17 00:00:00 2001
From: Fred Isaman <iisaman@netapp.com>
Date: Wed, 23 Mar 2011 13:27:45 +0000
Subject: NFSv4.1: rearrange nfs_commit_rpcsetup
Reorder nfs_commit_rpcsetup, preparing for a pnfs entry point.
Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/write.c | 60 ++++++++++++++++++++++++++++++++--------------------------
1 file changed, 33 insertions(+), 27 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 92b4a6614fd3..bf672faa4885 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -1293,32 +1293,49 @@ static void nfs_commitdata_release(void *data)
nfs_commit_free(wdata);
}
-/*
- * Set up the argument/result storage required for the RPC call.
- */
-static int nfs_commit_rpcsetup(struct list_head *head,
- struct nfs_write_data *data,
- int how)
+static int nfs_initiate_commit(struct nfs_write_data *data, struct rpc_clnt *clnt,
+ const struct rpc_call_ops *call_ops,
+ int how)
{
- struct nfs_page *first = nfs_list_entry(head->next);
- struct inode *inode = first->wb_context->path.dentry->d_inode;
- int priority = flush_task_priority(how);
struct rpc_task *task;
+ int priority = flush_task_priority(how);
struct rpc_message msg = {
.rpc_argp = &data->args,
.rpc_resp = &data->res,
- .rpc_cred = first->wb_context->cred,
+ .rpc_cred = data->cred,
};
struct rpc_task_setup task_setup_data = {
.task = &data->task,
- .rpc_client = NFS_CLIENT(inode),
+ .rpc_client = clnt,
.rpc_message = &msg,
- .callback_ops = &nfs_commit_ops,
+ .callback_ops = call_ops,
.callback_data = data,
.workqueue = nfsiod_workqueue,
.flags = RPC_TASK_ASYNC,
.priority = priority,
};
+ /* Set up the initial task struct. */
+ NFS_PROTO(data->inode)->commit_setup(data, &msg);
+
+ dprintk("NFS: %5u initiated commit call\n", data->task.tk_pid);
+
+ task = rpc_run_task(&task_setup_data);
+ if (IS_ERR(task))
+ return PTR_ERR(task);
+ if (how & FLUSH_SYNC)
+ rpc_wait_for_completion_task(task);
+ rpc_put_task(task);
+ return 0;
+}
+
+/*
+ * Set up the argument/result storage required for the RPC call.
+ */
+static void nfs_init_commit(struct nfs_write_data *data,
+ struct list_head *head)
+{
+ struct nfs_page *first = nfs_list_entry(head->next);
+ struct inode *inode = first->wb_context->path.dentry->d_inode;
/* Set up the RPC argument and reply structs
* NB: take care not to mess about with data->commit et al. */
@@ -1326,7 +1343,8 @@ static int nfs_commit_rpcsetup(struct list_head *head,
list_splice_init(head, &data->pages);
data->inode = inode;
- data->cred = msg.rpc_cred;
+ data->cred = first->wb_context->cred;
+ data->mds_ops = &nfs_commit_ops;
data->args.fh = NFS_FH(data->inode);
/* Note: we always request a commit of the entire inode */
@@ -1337,19 +1355,6 @@ static int nfs_commit_rpcsetup(struct list_head *head,
data->res.fattr = &data->fattr;
data->res.verf = &data->verf;
nfs_fattr_init(&data->fattr);
-
- /* Set up the initial task struct. */
- NFS_PROTO(inode)->commit_setup(data, &msg);
-
- dprintk("NFS: %5u initiated commit call\n", data->task.tk_pid);
-
- task = rpc_run_task(&task_setup_data);
- if (IS_ERR(task))
- return PTR_ERR(task);
- if (how & FLUSH_SYNC)
- rpc_wait_for_completion_task(task);
- rpc_put_task(task);
- return 0;
}
/*
@@ -1367,7 +1372,8 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how)
goto out_bad;
/* Set up the argument struct */
- return nfs_commit_rpcsetup(head, data, how);
+ nfs_init_commit(data, head);
+ return nfs_initiate_commit(data, NFS_CLIENT(inode), data->mds_ops, how);
out_bad:
while (!list_empty(head)) {
req = nfs_list_entry(head->next);
--
cgit v1.2.3
From 5f452431e229ea14df61b0ac89eed5881414eef9 Mon Sep 17 00:00:00 2001
From: Fred Isaman <iisaman@netapp.com>
Date: Wed, 23 Mar 2011 13:27:46 +0000
Subject: NFSv4.1: add callback to nfs4_commit_done
Add a callback that the pnfs layout driver can use to do its own
error handling of the data server's COMMIT response.
Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/nfs4proc.c | 14 ++++++++++----
1 file changed, 10 insertions(+), 4 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 1d84e7088af9..bcc29bdc9bee 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -3195,12 +3195,9 @@ static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_messag
msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
}
-static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
+static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
{
struct inode *inode = data->inode;
-
- if (!nfs4_sequence_done(task, &data->res.seq_res))
- return -EAGAIN;
if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
@@ -3210,11 +3207,20 @@ static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
return 0;
}
+static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
+{
+ if (!nfs4_sequence_done(task, &data->res.seq_res))
+ return -EAGAIN;
+ return data->write_done_cb(task, data);
+}
+
static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
{
struct nfs_server *server = NFS_SERVER(data->inode);
data->args.bitmask = server->cache_consistency_bitmask;
+ if (!data->write_done_cb)
+ data->write_done_cb = nfs4_commit_done_cb;
data->res.server = server;
msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
}
--
cgit v1.2.3
From 64bfeb49bd1c2351a8415f7fe6b25eef208872a5 Mon Sep 17 00:00:00 2001
From: Fred Isaman <iisaman@netapp.com>
Date: Wed, 23 Mar 2011 13:27:47 +0000
Subject: NFSv4.1: pull error handling out of nfs_commit_list
Create a separate support function for later use by data server
commit code.
Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/write.c | 26 ++++++++++++++++----------
1 file changed, 16 insertions(+), 10 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index bf672faa4885..dbc801810e75 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -1357,6 +1357,21 @@ static void nfs_init_commit(struct nfs_write_data *data,
nfs_fattr_init(&data->fattr);
}
+static void nfs_retry_commit(struct list_head *page_list)
+{
+ struct nfs_page *req;
+
+ while (!list_empty(page_list)) {
+ req = nfs_list_entry(page_list->next);
+ nfs_list_remove_request(req);
+ nfs_mark_request_commit(req);
+ dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
+ dec_bdi_stat(req->wb_page->mapping->backing_dev_info,
+ BDI_RECLAIMABLE);
+ nfs_clear_page_tag_locked(req);
+ }
+}
+
/*
* Commit dirty pages
*/
@@ -1364,7 +1379,6 @@ static int
nfs_commit_list(struct inode *inode, struct list_head *head, int how)
{
struct nfs_write_data *data;
- struct nfs_page *req;
data = nfs_commitdata_alloc();
@@ -1375,15 +1389,7 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how)
nfs_init_commit(data, head);
return nfs_initiate_commit(data, NFS_CLIENT(inode), data->mds_ops, how);
out_bad:
- while (!list_empty(head)) {
- req = nfs_list_entry(head->next);
- nfs_list_remove_request(req);
- nfs_mark_request_commit(req);
- dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
- dec_bdi_stat(req->wb_page->mapping->backing_dev_info,
- BDI_RECLAIMABLE);
- nfs_clear_page_tag_locked(req);
- }
+ nfs_retry_commit(head);
nfs_commit_clear_lock(NFS_I(inode));
return -ENOMEM;
}
--
cgit v1.2.3
From 5917ce8440ba0b3f2adee613b5f1258ac5efff02 Mon Sep 17 00:00:00 2001
From: Fred Isaman <iisaman@netapp.com>
Date: Wed, 23 Mar 2011 13:27:48 +0000
Subject: NFSv4.1: pull out code from nfs_commit_release
Create a separate support function for later use by data server
commit code.
Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/write.c | 12 +++++++++---
1 file changed, 9 insertions(+), 3 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index dbc801810e75..f5f005e9db48 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -1409,10 +1409,9 @@ static void nfs_commit_done(struct rpc_task *task, void *calldata)
return;
}
-static void nfs_commit_release(void *calldata)
+static void nfs_commit_release_pages(struct nfs_write_data *data)
{
- struct nfs_write_data *data = calldata;
- struct nfs_page *req;
+ struct nfs_page *req;
int status = data->task.tk_status;
while (!list_empty(&data->pages)) {
@@ -1446,6 +1445,13 @@ static void nfs_commit_release(void *calldata)
next:
nfs_clear_page_tag_locked(req);
}
+}
+
+static void nfs_commit_release(void *calldata)
+{
+ struct nfs_write_data *data = calldata;
+
+ nfs_commit_release_pages(data);
nfs_commit_clear_lock(NFS_I(data->inode));
nfs_commitdata_release(calldata);
}
--
cgit v1.2.3
From c879513e91c9e38957f0d9647d22469f1887e4da Mon Sep 17 00:00:00 2001
From: Fred Isaman <iisaman@netapp.com>
Date: Wed, 23 Mar 2011 13:27:49 +0000
Subject: NFSv4.1: shift filelayout_free_lseg
Move it up to avoid forward declaration in later patch.
Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/nfs4filelayout.c | 20 ++++++++++----------
1 file changed, 10 insertions(+), 10 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c
index 428558464817..9401afd5650e 100644
--- a/fs/nfs/nfs4filelayout.c
+++ b/fs/nfs/nfs4filelayout.c
@@ -496,6 +496,16 @@ filelayout_decode_layout(struct pnfs_layout_hdr *flo,
return 0;
}
+static void
+filelayout_free_lseg(struct pnfs_layout_segment *lseg)
+{
+ struct nfs4_filelayout_segment *fl = FILELAYOUT_LSEG(lseg);
+
+ dprintk("--> %s\n", __func__);
+ nfs4_fl_put_deviceid(fl->dsaddr);
+ _filelayout_free_lseg(fl);
+}
+
static struct pnfs_layout_segment *
filelayout_alloc_lseg(struct pnfs_layout_hdr *layoutid,
struct nfs4_layoutget_res *lgr)
@@ -517,16 +527,6 @@ filelayout_alloc_lseg(struct pnfs_layout_hdr *layoutid,
return &fl->generic_hdr;
}
-static void
-filelayout_free_lseg(struct pnfs_layout_segment *lseg)
-{
- struct nfs4_filelayout_segment *fl = FILELAYOUT_LSEG(lseg);
-
- dprintk("--> %s\n", __func__);
- nfs4_fl_put_deviceid(fl->dsaddr);
- _filelayout_free_lseg(fl);
-}
-
/*
* filelayout_pg_test(). Called by nfs_can_coalesce_requests()
*
--
cgit v1.2.3
From 425eb736cd905181a4dd4dc8d66342a7c7ab2f27 Mon Sep 17 00:00:00 2001
From: Fred Isaman <iisaman@netapp.com>
Date: Wed, 23 Mar 2011 13:27:50 +0000
Subject: NFSv4.1: alloc and free commit_buckets
Create a preallocated list header to hold nfs_pages for each
non-MDS COMMIT destination. Note this is not necessarily each DS,
but is basically each <DS, fh> pair.
Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/nfs4filelayout.c | 22 ++++++++++++++++++++++
fs/nfs/nfs4filelayout.h | 2 ++
2 files changed, 24 insertions(+)
(limited to 'fs')
diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c
index 9401afd5650e..03ff80c67c6e 100644
--- a/fs/nfs/nfs4filelayout.c
+++ b/fs/nfs/nfs4filelayout.c
@@ -503,6 +503,7 @@ filelayout_free_lseg(struct pnfs_layout_segment *lseg)
dprintk("--> %s\n", __func__);
nfs4_fl_put_deviceid(fl->dsaddr);
+ kfree(fl->commit_buckets);
_filelayout_free_lseg(fl);
}
@@ -524,6 +525,27 @@ filelayout_alloc_lseg(struct pnfs_layout_hdr *layoutid,
_filelayout_free_lseg(fl);
return NULL;
}
+
+ /* This assumes there is only one IOMODE_RW lseg. What
+ * we really want to do is have a layout_hdr level
+ * dictionary of <multipath_list4, fh> keys, each
+ * associated with a struct list_head, populated by calls
+ * to filelayout_write_pagelist().
+ * */
+ if ((!fl->commit_through_mds) && (lgr->range.iomode == IOMODE_RW)) {
+ int i;
+ int size = (fl->stripe_type == STRIPE_SPARSE) ?
+ fl->dsaddr->ds_num : fl->dsaddr->stripe_count;
+
+ fl->commit_buckets = kcalloc(size, sizeof(struct list_head), GFP_KERNEL);
+ if (!fl->commit_buckets) {
+ filelayout_free_lseg(&fl->generic_hdr);
+ return NULL;
+ }
+ fl->number_of_buckets = size;
+ for (i = 0; i < size; i++)
+ INIT_LIST_HEAD(&fl->commit_buckets[i]);
+ }
return &fl->generic_hdr;
}
diff --git a/fs/nfs/nfs4filelayout.h b/fs/nfs/nfs4filelayout.h
index ee0c907742b5..085a354e0f08 100644
--- a/fs/nfs/nfs4filelayout.h
+++ b/fs/nfs/nfs4filelayout.h
@@ -79,6 +79,8 @@ struct nfs4_filelayout_segment {
struct nfs4_file_layout_dsaddr *dsaddr; /* Point to GETDEVINFO data */
unsigned int num_fh;
struct nfs_fh **fh_array;
+ struct list_head *commit_buckets; /* Sort commits to ds */
+ int number_of_buckets;
};
static inline struct nfs4_filelayout_segment *
--
cgit v1.2.3
From a861a1e1c398fe34701569fd8ac9225dfe0a9a7e Mon Sep 17 00:00:00 2001
From: Fred Isaman <iisaman@netapp.com>
Date: Wed, 23 Mar 2011 13:27:51 +0000
Subject: NFSv4.1: add generic layer hooks for pnfs COMMIT
We create three major hooks for the pnfs code.
pnfs_mark_request_commit() is called during writeback_done from
nfs_mark_request_commit, which gives the driver an opportunity to
claim it wants control over commiting a particular req.
pnfs_choose_commit_list() is called from nfs_scan_list
to choose which list a given req should be added to, based on
where we intend to send it for COMMIT. It is up to the driver
to have preallocated list headers for each destination it may need.
pnfs_commit_list() is how the driver actually takes control, it is
used instead of nfs_commit_list().
In order to pass information between the above functions, we create
a union in nfs_page to hold a lseg (which is possible because the req is
not on any list while in transition), and add some flags to indicate
if we need to use the pnfs code.
Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/pagelist.c | 5 ++--
fs/nfs/pnfs.h | 73 ++++++++++++++++++++++++++++++++++++++++++++++++
fs/nfs/write.c | 41 +++++++++++++++++----------
include/linux/nfs_fs.h | 1 +
include/linux/nfs_page.h | 6 +++-
5 files changed, 108 insertions(+), 18 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c
index fd85618149a1..87a593c2b055 100644
--- a/fs/nfs/pagelist.c
+++ b/fs/nfs/pagelist.c
@@ -398,6 +398,7 @@ int nfs_scan_list(struct nfs_inode *nfsi,
pgoff_t idx_end;
int found, i;
int res;
+ struct list_head *list;
res = 0;
if (npages == 0)
@@ -418,10 +419,10 @@ int nfs_scan_list(struct nfs_inode *nfsi,
idx_start = req->wb_index + 1;
if (nfs_set_page_tag_locked(req)) {
kref_get(&req->wb_kref);
- nfs_list_remove_request(req);
radix_tree_tag_clear(&nfsi->nfs_page_tree,
req->wb_index, tag);
- nfs_list_add_request(req, dst);
+ list = pnfs_choose_commit_list(req, dst);
+ nfs_list_add_request(req, list);
res++;
if (res == INT_MAX)
goto out;
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index 6380b9405bcd..5370f1b9aa43 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -74,6 +74,13 @@ struct pnfs_layoutdriver_type {
/* test for nfs page cache coalescing */
int (*pg_test)(struct nfs_pageio_descriptor *, struct nfs_page *, struct nfs_page *);
+ /* Returns true if layoutdriver wants to divert this request to
+ * driver's commit routine.
+ */
+ bool (*mark_pnfs_commit)(struct pnfs_layout_segment *lseg);
+ struct list_head * (*choose_commit_list) (struct nfs_page *req);
+ int (*commit_pagelist)(struct inode *inode, struct list_head *mds_pages, int how);
+
/*
* Return PNFS_ATTEMPTED to indicate the layout code has attempted
* I/O, else return PNFS_NOT_ATTEMPTED to fall back to normal NFS
@@ -169,6 +176,51 @@ static inline int pnfs_enabled_sb(struct nfs_server *nfss)
return nfss->pnfs_curr_ld != NULL;
}
+static inline void
+pnfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg)
+{
+ if (lseg) {
+ struct pnfs_layoutdriver_type *ld;
+
+ ld = NFS_SERVER(req->wb_page->mapping->host)->pnfs_curr_ld;
+ if (ld->mark_pnfs_commit && ld->mark_pnfs_commit(lseg)) {
+ set_bit(PG_PNFS_COMMIT, &req->wb_flags);
+ req->wb_commit_lseg = get_lseg(lseg);
+ }
+ }
+}
+
+static inline int
+pnfs_commit_list(struct inode *inode, struct list_head *mds_pages, int how)
+{
+ if (!test_and_clear_bit(NFS_INO_PNFS_COMMIT, &NFS_I(inode)->flags))
+ return PNFS_NOT_ATTEMPTED;
+ return NFS_SERVER(inode)->pnfs_curr_ld->commit_pagelist(inode, mds_pages, how);
+}
+
+static inline struct list_head *
+pnfs_choose_commit_list(struct nfs_page *req, struct list_head *mds)
+{
+ struct list_head *rv;
+
+ if (test_and_clear_bit(PG_PNFS_COMMIT, &req->wb_flags)) {
+ struct inode *inode = req->wb_commit_lseg->pls_layout->plh_inode;
+
+ set_bit(NFS_INO_PNFS_COMMIT, &NFS_I(inode)->flags);
+ rv = NFS_SERVER(inode)->pnfs_curr_ld->choose_commit_list(req);
+ /* matched by ref taken when PG_PNFS_COMMIT is set */
+ put_lseg(req->wb_commit_lseg);
+ } else
+ rv = mds;
+ return rv;
+}
+
+static inline void pnfs_clear_request_commit(struct nfs_page *req)
+{
+ if (test_and_clear_bit(PG_PNFS_COMMIT, &req->wb_flags))
+ put_lseg(req->wb_commit_lseg);
+}
+
#else /* CONFIG_NFS_V4_1 */
static inline void pnfs_destroy_all_layouts(struct nfs_client *clp)
@@ -252,6 +304,27 @@ pnfs_pageio_init_write(struct nfs_pageio_descriptor *pgio, struct inode *ino)
pgio->pg_test = NULL;
}
+static inline void
+pnfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg)
+{
+}
+
+static inline int
+pnfs_commit_list(struct inode *inode, struct list_head *mds_pages, int how)
+{
+ return PNFS_NOT_ATTEMPTED;
+}
+
+static inline struct list_head *
+pnfs_choose_commit_list(struct nfs_page *req, struct list_head *mds)
+{
+ return mds;
+}
+
+static inline void pnfs_clear_request_commit(struct nfs_page *req)
+{
+}
+
#endif /* CONFIG_NFS_V4_1 */
#endif /* FS_NFS_PNFS_H */
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index f5f005e9db48..6927a18b6891 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -441,7 +441,7 @@ nfs_mark_request_dirty(struct nfs_page *req)
* Add a request to the inode's commit list.
*/
static void
-nfs_mark_request_commit(struct nfs_page *req)
+nfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg)
{
struct inode *inode = req->wb_context->path.dentry->d_inode;
struct nfs_inode *nfsi = NFS_I(inode);
@@ -453,6 +453,7 @@ nfs_mark_request_commit(struct nfs_page *req)
NFS_PAGE_TAG_COMMIT);
nfsi->ncommit++;
spin_unlock(&inode->i_lock);
+ pnfs_mark_request_commit(req, lseg);
inc_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
inc_bdi_stat(req->wb_page->mapping->backing_dev_info, BDI_RECLAIMABLE);
__mark_inode_dirty(inode, I_DIRTY_DATASYNC);
@@ -481,10 +482,11 @@ int nfs_write_need_commit(struct nfs_write_data *data)
}
static inline
-int nfs_reschedule_unstable_write(struct nfs_page *req)
+int nfs_reschedule_unstable_write(struct nfs_page *req,
+ struct nfs_write_data *data)
{
if (test_and_clear_bit(PG_NEED_COMMIT, &req->wb_flags)) {
- nfs_mark_request_commit(req);
+ nfs_mark_request_commit(req, data->lseg);
return 1;
}
if (test_and_clear_bit(PG_NEED_RESCHED, &req->wb_flags)) {
@@ -495,7 +497,7 @@ int nfs_reschedule_unstable_write(struct nfs_page *req)
}
#else
static inline void
-nfs_mark_request_commit(struct nfs_page *req)
+nfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg)
{
}
@@ -512,7 +514,8 @@ int nfs_write_need_commit(struct nfs_write_data *data)
}
static inline
-int nfs_reschedule_unstable_write(struct nfs_page *req)
+int nfs_reschedule_unstable_write(struct nfs_page *req,
+ struct nfs_write_data *data)
{
return 0;
}
@@ -615,9 +618,11 @@ static struct nfs_page *nfs_try_to_update_request(struct inode *inode,
}
if (nfs_clear_request_commit(req) &&
- radix_tree_tag_clear(&NFS_I(inode)->nfs_page_tree,
- req->wb_index, NFS_PAGE_TAG_COMMIT) != NULL)
+ radix_tree_tag_clear(&NFS_I(inode)->nfs_page_tree,
+ req->wb_index, NFS_PAGE_TAG_COMMIT) != NULL) {
NFS_I(inode)->ncommit--;
+ pnfs_clear_request_commit(req);
+ }
/* Okay, the request matches. Update the region */
if (offset < req->wb_offset) {
@@ -765,11 +770,12 @@ int nfs_updatepage(struct file *file, struct page *page,
return status;
}
-static void nfs_writepage_release(struct nfs_page *req)
+static void nfs_writepage_release(struct nfs_page *req,
+ struct nfs_write_data *data)
{
struct page *page = req->wb_page;
- if (PageError(req->wb_page) || !nfs_reschedule_unstable_write(req))
+ if (PageError(req->wb_page) || !nfs_reschedule_unstable_write(req, data))
nfs_inode_remove_request(req);
nfs_clear_page_tag_locked(req);
nfs_end_page_writeback(page);
@@ -1087,7 +1093,7 @@ static void nfs_writeback_release_partial(void *calldata)
out:
if (atomic_dec_and_test(&req->wb_complete))
- nfs_writepage_release(req);
+ nfs_writepage_release(req, data);
nfs_writedata_release(calldata);
}
@@ -1154,7 +1160,7 @@ static void nfs_writeback_release_full(void *calldata)
if (nfs_write_need_commit(data)) {
memcpy(&req->wb_verf, &data->verf, sizeof(req->wb_verf));
- nfs_mark_request_commit(req);
+ nfs_mark_request_commit(req, data->lseg);
dprintk(" marked for commit\n");
goto next;
}
@@ -1357,14 +1363,15 @@ static void nfs_init_commit(struct nfs_write_data *data,
nfs_fattr_init(&data->fattr);
}
-static void nfs_retry_commit(struct list_head *page_list)
+static void nfs_retry_commit(struct list_head *page_list,
+ struct pnfs_layout_segment *lseg)
{
struct nfs_page *req;
while (!list_empty(page_list)) {
req = nfs_list_entry(page_list->next);
nfs_list_remove_request(req);
- nfs_mark_request_commit(req);
+ nfs_mark_request_commit(req, lseg);
dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
dec_bdi_stat(req->wb_page->mapping->backing_dev_info,
BDI_RECLAIMABLE);
@@ -1389,7 +1396,7 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how)
nfs_init_commit(data, head);
return nfs_initiate_commit(data, NFS_CLIENT(inode), data->mds_ops, how);
out_bad:
- nfs_retry_commit(head);
+ nfs_retry_commit(head, NULL);
nfs_commit_clear_lock(NFS_I(inode));
return -ENOMEM;
}
@@ -1477,7 +1484,11 @@ int nfs_commit_inode(struct inode *inode, int how)
res = nfs_scan_commit(inode, &head, 0, 0);
spin_unlock(&inode->i_lock);
if (res) {
- int error = nfs_commit_list(inode, &head, how);
+ int error;
+
+ error = pnfs_commit_list(inode, &head, how);
+ if (error == PNFS_NOT_ATTEMPTED)
+ error = nfs_commit_list(inode, &head, how);
if (error < 0)
return error;
if (!may_wait)
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index 4179c368844b..eddda6ce7c42 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -226,6 +226,7 @@ struct nfs_inode {
#define NFS_INO_FSCACHE (5) /* inode can be cached by FS-Cache */
#define NFS_INO_FSCACHE_LOCK (6) /* FS-Cache cookie management lock */
#define NFS_INO_COMMIT (7) /* inode is committing unstable writes */
+#define NFS_INO_PNFS_COMMIT (8) /* use pnfs code for commit */
static inline struct nfs_inode *NFS_I(const struct inode *inode)
{
diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h
index 92d54c81f51e..8023e4e25133 100644
--- a/include/linux/nfs_page.h
+++ b/include/linux/nfs_page.h
@@ -33,11 +33,15 @@ enum {
PG_CLEAN,
PG_NEED_COMMIT,
PG_NEED_RESCHED,
+ PG_PNFS_COMMIT,
};
struct nfs_inode;
struct nfs_page {
- struct list_head wb_list; /* Defines state of page: */
+ union {
+ struct list_head wb_list; /* Defines state of page: */
+ struct pnfs_layout_segment *wb_commit_lseg; /* Used when PG_PNFS_COMMIT set */
+ };
struct page *wb_page; /* page to read in/write out */
struct nfs_open_context *wb_context; /* File state context info */
struct nfs_lock_context *wb_lock_context; /* lock context info */
--
cgit v1.2.3
From 988b6dceb0ae6d642587c8594529b94f6be0c5ea Mon Sep 17 00:00:00 2001
From: Fred Isaman <iisaman@netapp.com>
Date: Wed, 23 Mar 2011 13:27:52 +0000
Subject: NFSv4.1: remove GETATTR from ds commits
Any COMMIT compound directed to a data server needs to have the
GETATTR calls suppressed. We here, make sure the field we are testing
(data->lseg) is set and refcounted correctly.
Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/nfs4proc.c | 8 ++++++--
fs/nfs/nfs4xdr.c | 8 +++++---
fs/nfs/write.c | 7 +++++--
3 files changed, 16 insertions(+), 7 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index bcc29bdc9bee..5d61cccc8d4d 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -3217,8 +3217,12 @@ static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
{
struct nfs_server *server = NFS_SERVER(data->inode);
-
- data->args.bitmask = server->cache_consistency_bitmask;
+
+ if (data->lseg) {
+ data->args.bitmask = NULL;
+ data->res.fattr = NULL;
+ } else
+ data->args.bitmask = server->cache_consistency_bitmask;
if (!data->write_done_cb)
data->write_done_cb = nfs4_commit_done_cb;
data->res.server = server;
diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
index 0cf560f77884..07cdf925c524 100644
--- a/fs/nfs/nfs4xdr.c
+++ b/fs/nfs/nfs4xdr.c
@@ -2294,7 +2294,8 @@ static void nfs4_xdr_enc_commit(struct rpc_rqst *req, struct xdr_stream *xdr,
encode_sequence(xdr, &args->seq_args, &hdr);
encode_putfh(xdr, args->fh, &hdr);
encode_commit(xdr, args, &hdr);
- encode_getfattr(xdr, args->bitmask, &hdr);
+ if (args->bitmask)
+ encode_getfattr(xdr, args->bitmask, &hdr);
encode_nops(&hdr);
}
@@ -5723,8 +5724,9 @@ static int nfs4_xdr_dec_commit(struct rpc_rqst *rqstp, struct xdr_stream *xdr,
status = decode_commit(xdr, res);
if (status)
goto out;
- decode_getfattr(xdr, res->fattr, res->server,
- !RPC_IS_ASYNC(rqstp->rq_task));
+ if (res->fattr)
+ decode_getfattr(xdr, res->fattr, res->server,
+ !RPC_IS_ASYNC(rqstp->rq_task));
out:
return status;
}
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 6927a18b6891..cae5d160d835 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -1295,6 +1295,7 @@ static void nfs_commitdata_release(void *data)
{
struct nfs_write_data *wdata = data;
+ put_lseg(wdata->lseg);
put_nfs_open_context(wdata->args.context);
nfs_commit_free(wdata);
}
@@ -1338,7 +1339,8 @@ static int nfs_initiate_commit(struct nfs_write_data *data, struct rpc_clnt *cln
* Set up the argument/result storage required for the RPC call.
*/
static void nfs_init_commit(struct nfs_write_data *data,
- struct list_head *head)
+ struct list_head *head,
+ struct pnfs_layout_segment *lseg)
{
struct nfs_page *first = nfs_list_entry(head->next);
struct inode *inode = first->wb_context->path.dentry->d_inode;
@@ -1350,6 +1352,7 @@ static void nfs_init_commit(struct nfs_write_data *data,
data->inode = inode;
data->cred = first->wb_context->cred;
+ data->lseg = lseg; /* reference transferred */
data->mds_ops = &nfs_commit_ops;
data->args.fh = NFS_FH(data->inode);
@@ -1393,7 +1396,7 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how)
goto out_bad;
/* Set up the argument struct */
- nfs_init_commit(data, head);
+ nfs_init_commit(data, head, NULL);
return nfs_initiate_commit(data, NFS_CLIENT(inode), data->mds_ops, how);
out_bad:
nfs_retry_commit(head, NULL);
--
cgit v1.2.3
From e0c2b3801828aadb65dec9f67f7c6b7a675ad007 Mon Sep 17 00:00:00 2001
From: Fred Isaman <iisaman@netapp.com>
Date: Wed, 23 Mar 2011 13:27:53 +0000
Subject: NFSv4.1: filelayout driver specific code for COMMIT
Implement all the hooks created in the previous patches.
This requires exporting quite a few functions and adding a few
structure fields.
Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/inode.c | 1 +
fs/nfs/internal.h | 14 +++
fs/nfs/nfs4filelayout.c | 235 ++++++++++++++++++++++++++++++++++++++++++++++++
fs/nfs/pnfs.c | 1 +
fs/nfs/write.c | 21 +++--
include/linux/nfs_fs.h | 1 +
include/linux/nfs_xdr.h | 1 +
7 files changed, 267 insertions(+), 7 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index 477a2e512b39..229e586b1a20 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -1470,6 +1470,7 @@ static inline void nfs4_init_once(struct nfs_inode *nfsi)
nfsi->delegation_state = 0;
init_rwsem(&nfsi->rwsem);
nfsi->layout = NULL;
+ atomic_set(&nfsi->commits_outstanding, 0);
#endif
}
diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
index d1ddc23c404d..708705062216 100644
--- a/fs/nfs/internal.h
+++ b/fs/nfs/internal.h
@@ -276,11 +276,25 @@ extern int nfs_initiate_read(struct nfs_read_data *data, struct rpc_clnt *clnt,
extern void nfs_read_prepare(struct rpc_task *task, void *calldata);
/* write.c */
+extern void nfs_commit_free(struct nfs_write_data *p);
extern int nfs_initiate_write(struct nfs_write_data *data,
struct rpc_clnt *clnt,
const struct rpc_call_ops *call_ops,
int how);
extern void nfs_write_prepare(struct rpc_task *task, void *calldata);
+extern int nfs_initiate_commit(struct nfs_write_data *data,
+ struct rpc_clnt *clnt,
+ const struct rpc_call_ops *call_ops,
+ int how);
+extern void nfs_init_commit(struct nfs_write_data *data,
+ struct list_head *head,
+ struct pnfs_layout_segment *lseg);
+void nfs_retry_commit(struct list_head *page_list,
+ struct pnfs_layout_segment *lseg);
+void nfs_commit_clear_lock(struct nfs_inode *nfsi);
+void nfs_commitdata_release(void *data);
+void nfs_commit_release_pages(struct nfs_write_data *data);
+
#ifdef CONFIG_MIGRATION
extern int nfs_migrate_page(struct address_space *,
struct page *, struct page *);
diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c
index 03ff80c67c6e..97e75a22af72 100644
--- a/fs/nfs/nfs4filelayout.c
+++ b/fs/nfs/nfs4filelayout.c
@@ -213,6 +213,37 @@ static int filelayout_write_done_cb(struct rpc_task *task,
return 0;
}
+/* Fake up some data that will cause nfs_commit_release to retry the writes. */
+static void prepare_to_resend_writes(struct nfs_write_data *data)
+{
+ struct nfs_page *first = nfs_list_entry(data->pages.next);
+
+ data->task.tk_status = 0;
+ memcpy(data->verf.verifier, first->wb_verf.verifier,
+ sizeof(first->wb_verf.verifier));
+ data->verf.verifier[0]++; /* ensure verifier mismatch */
+}
+
+static int filelayout_commit_done_cb(struct rpc_task *task,
+ struct nfs_write_data *data)
+{
+ int reset = 0;
+
+ if (filelayout_async_handle_error(task, data->args.context->state,
+ data->ds_clp, &reset) == -EAGAIN) {
+ dprintk("%s calling restart ds_clp %p ds_clp->cl_session %p\n",
+ __func__, data->ds_clp, data->ds_clp->cl_session);
+ if (reset) {
+ prepare_to_resend_writes(data);
+ filelayout_set_lo_fail(data->lseg);
+ } else
+ nfs_restart_rpc(task, data->ds_clp);
+ return -EAGAIN;
+ }
+
+ return 0;
+}
+
static void filelayout_write_prepare(struct rpc_task *task, void *data)
{
struct nfs_write_data *wdata = (struct nfs_write_data *)data;
@@ -240,6 +271,16 @@ static void filelayout_write_release(void *data)
wdata->mds_ops->rpc_release(data);
}
+static void filelayout_commit_release(void *data)
+{
+ struct nfs_write_data *wdata = (struct nfs_write_data *)data;
+
+ nfs_commit_release_pages(wdata);
+ if (atomic_dec_and_test(&NFS_I(wdata->inode)->commits_outstanding))
+ nfs_commit_clear_lock(NFS_I(wdata->inode));
+ nfs_commitdata_release(wdata);
+}
+
struct rpc_call_ops filelayout_read_call_ops = {
.rpc_call_prepare = filelayout_read_prepare,
.rpc_call_done = filelayout_read_call_done,
@@ -252,6 +293,12 @@ struct rpc_call_ops filelayout_write_call_ops = {
.rpc_release = filelayout_write_release,
};
+struct rpc_call_ops filelayout_commit_call_ops = {
+ .rpc_call_prepare = filelayout_write_prepare,
+ .rpc_call_done = filelayout_write_call_done,
+ .rpc_release = filelayout_commit_release,
+};
+
static enum pnfs_try_status
filelayout_read_pagelist(struct nfs_read_data *data)
{
@@ -574,6 +621,191 @@ filelayout_pg_test(struct nfs_pageio_descriptor *pgio, struct nfs_page *prev,
return (p_stripe == r_stripe);
}
+static bool filelayout_mark_pnfs_commit(struct pnfs_layout_segment *lseg)
+{
+ return !FILELAYOUT_LSEG(lseg)->commit_through_mds;
+}
+
+static u32 select_bucket_index(struct nfs4_filelayout_segment *fl, u32 j)
+{
+ if (fl->stripe_type == STRIPE_SPARSE)
+ return nfs4_fl_calc_ds_index(&fl->generic_hdr, j);
+ else
+ return j;
+}
+
+struct list_head *filelayout_choose_commit_list(struct nfs_page *req)
+{
+ struct pnfs_layout_segment *lseg = req->wb_commit_lseg;
+ struct nfs4_filelayout_segment *fl = FILELAYOUT_LSEG(lseg);
+ u32 i, j;
+ struct list_head *list;
+
+ /* Note that we are calling nfs4_fl_calc_j_index on each page
+ * that ends up being committed to a data server. An attractive
+ * alternative is to add a field to nfs_write_data and nfs_page
+ * to store the value calculated in filelayout_write_pagelist
+ * and just use that here.
+ */
+ j = nfs4_fl_calc_j_index(lseg,
+ (loff_t)req->wb_index << PAGE_CACHE_SHIFT);
+ i = select_bucket_index(fl, j);
+ list = &fl->commit_buckets[i];
+ if (list_empty(list)) {
+ /* Non-empty buckets hold a reference on the lseg */
+ get_lseg(lseg);
+ }
+ return list;
+}
+
+static u32 calc_ds_index_from_commit(struct pnfs_layout_segment *lseg, u32 i)
+{
+ struct nfs4_filelayout_segment *flseg = FILELAYOUT_LSEG(lseg);
+
+ if (flseg->stripe_type == STRIPE_SPARSE)
+ return i;
+ else
+ return nfs4_fl_calc_ds_index(lseg, i);
+}
+
+static struct nfs_fh *
+select_ds_fh_from_commit(struct pnfs_layout_segment *lseg, u32 i)
+{
+ struct nfs4_filelayout_segment *flseg = FILELAYOUT_LSEG(lseg);
+
+ if (flseg->stripe_type == STRIPE_SPARSE) {
+ if (flseg->num_fh == 1)
+ i = 0;
+ else if (flseg->num_fh == 0)
+ /* Use the MDS OPEN fh set in nfs_read_rpcsetup */
+ return NULL;
+ }
+ return flseg->fh_array[i];
+}
+
+static int filelayout_initiate_commit(struct nfs_write_data *data, int how)
+{
+ struct pnfs_layout_segment *lseg = data->lseg;
+ struct nfs4_pnfs_ds *ds;
+ u32 idx;
+ struct nfs_fh *fh;
+
+ idx = calc_ds_index_from_commit(lseg, data->ds_commit_index);
+ ds = nfs4_fl_prepare_ds(lseg, idx);
+ if (!ds) {
+ printk(KERN_ERR "%s: prepare_ds failed, use MDS\n", __func__);
+ set_bit(lo_fail_bit(IOMODE_RW), &lseg->pls_layout->plh_flags);
+ set_bit(lo_fail_bit(IOMODE_READ), &lseg->pls_layout->plh_flags);
+ prepare_to_resend_writes(data);
+ data->mds_ops->rpc_release(data);
+ return -EAGAIN;
+ }
+ dprintk("%s ino %lu, how %d\n", __func__, data->inode->i_ino, how);
+ data->write_done_cb = filelayout_commit_done_cb;
+ data->ds_clp = ds->ds_clp;
+ fh = select_ds_fh_from_commit(lseg, data->ds_commit_index);
+ if (fh)
+ data->args.fh = fh;
+ return nfs_initiate_commit(data, ds->ds_clp->cl_rpcclient,
+ &filelayout_commit_call_ops, how);
+}
+
+/*
+ * This is only useful while we are using whole file layouts.
+ */
+static struct pnfs_layout_segment *find_only_write_lseg(struct inode *inode)
+{
+ struct pnfs_layout_segment *lseg, *rv = NULL;
+
+ spin_lock(&inode->i_lock);
+ list_for_each_entry(lseg, &NFS_I(inode)->layout->plh_segs, pls_list)
+ if (lseg->pls_range.iomode == IOMODE_RW)
+ rv = get_lseg(lseg);
+ spin_unlock(&inode->i_lock);
+ return rv;
+}
+
+static int alloc_ds_commits(struct inode *inode, struct list_head *list)
+{
+ struct pnfs_layout_segment *lseg;
+ struct nfs4_filelayout_segment *fl;
+ struct nfs_write_data *data;
+ int i, j;
+
+ /* Won't need this when non-whole file layout segments are supported
+ * instead we will use a pnfs_layout_hdr structure */
+ lseg = find_only_write_lseg(inode);
+ if (!lseg)
+ return 0;
+ fl = FILELAYOUT_LSEG(lseg);
+ for (i = 0; i < fl->number_of_buckets; i++) {
+ if (list_empty(&fl->commit_buckets[i]))
+ continue;
+ data = nfs_commitdata_alloc();
+ if (!data)
+ goto out_bad;
+ data->ds_commit_index = i;
+ data->lseg = lseg;
+ list_add(&data->pages, list);
+ }
+ put_lseg(lseg);
+ return 0;
+
+out_bad:
+ for (j = i; j < fl->number_of_buckets; j++) {
+ if (list_empty(&fl->commit_buckets[i]))
+ continue;
+ nfs_retry_commit(&fl->commit_buckets[i], lseg);
+ put_lseg(lseg); /* associated with emptying bucket */
+ }
+ put_lseg(lseg);
+ /* Caller will clean up entries put on list */
+ return -ENOMEM;
+}
+
+/* This follows nfs_commit_list pretty closely */
+static int
+filelayout_commit_pagelist(struct inode *inode, struct list_head *mds_pages,
+ int how)
+{
+ struct nfs_write_data *data, *tmp;
+ LIST_HEAD(list);
+
+ if (!list_empty(mds_pages)) {
+ data = nfs_commitdata_alloc();
+ if (!data)
+ goto out_bad;
+ data->lseg = NULL;
+ list_add(&data->pages, &list);
+ }
+
+ if (alloc_ds_commits(inode, &list))
+ goto out_bad;
+
+ list_for_each_entry_safe(data, tmp, &list, pages) {
+ list_del_init(&data->pages);
+ atomic_inc(&NFS_I(inode)->commits_outstanding);
+ if (!data->lseg) {
+ nfs_init_commit(data, mds_pages, NULL);
+ nfs_initiate_commit(data, NFS_CLIENT(inode),
+ data->mds_ops, how);
+ } else {
+ nfs_init_commit(data, &FILELAYOUT_LSEG(data->lseg)->commit_buckets[data->ds_commit_index], data->lseg);
+ filelayout_initiate_commit(data, how);
+ }
+ }
+ return 0;
+ out_bad:
+ list_for_each_entry_safe(data, tmp, &list, pages) {
+ nfs_retry_commit(&data->pages, data->lseg);
+ list_del_init(&data->pages);
+ nfs_commit_free(data);
+ }
+ nfs_retry_commit(mds_pages, NULL);
+ nfs_commit_clear_lock(NFS_I(inode));
+ return -ENOMEM;
+}
+
static struct pnfs_layoutdriver_type filelayout_type = {
.id = LAYOUT_NFSV4_1_FILES,
.name = "LAYOUT_NFSV4_1_FILES",
@@ -581,6 +813,9 @@ static struct pnfs_layoutdriver_type filelayout_type = {
.alloc_lseg = filelayout_alloc_lseg,
.free_lseg = filelayout_free_lseg,
.pg_test = filelayout_pg_test,
+ .mark_pnfs_commit = filelayout_mark_pnfs_commit,
+ .choose_commit_list = filelayout_choose_commit_list,
+ .commit_pagelist = filelayout_commit_pagelist,
.read_pagelist = filelayout_read_pagelist,
.write_pagelist = filelayout_write_pagelist,
};
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index f38813a0a295..c67565965f2a 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -259,6 +259,7 @@ put_lseg(struct pnfs_layout_segment *lseg)
pnfs_free_lseg_list(&free_me);
}
}
+EXPORT_SYMBOL_GPL(put_lseg);
static bool
should_free_lseg(u32 lseg_iomode, u32 recall_iomode)
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index cae5d160d835..e7aeda0663c5 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -59,6 +59,7 @@ struct nfs_write_data *nfs_commitdata_alloc(void)
}
return p;
}
+EXPORT_SYMBOL_GPL(nfs_commitdata_alloc);
void nfs_commit_free(struct nfs_write_data *p)
{
@@ -66,6 +67,7 @@ void nfs_commit_free(struct nfs_write_data *p)
kfree(p->pagevec);
mempool_free(p, nfs_commit_mempool);
}
+EXPORT_SYMBOL_GPL(nfs_commit_free);
struct nfs_write_data *nfs_writedata_alloc(unsigned int pagecount)
{
@@ -1283,15 +1285,15 @@ static int nfs_commit_set_lock(struct nfs_inode *nfsi, int may_wait)
return (ret < 0) ? ret : 1;
}
-static void nfs_commit_clear_lock(struct nfs_inode *nfsi)
+void nfs_commit_clear_lock(struct nfs_inode *nfsi)
{
clear_bit(NFS_INO_COMMIT, &nfsi->flags);
smp_mb__after_clear_bit();
wake_up_bit(&nfsi->flags, NFS_INO_COMMIT);
}
+EXPORT_SYMBOL_GPL(nfs_commit_clear_lock);
-
-static void nfs_commitdata_release(void *data)
+void nfs_commitdata_release(void *data)
{
struct nfs_write_data *wdata = data;
@@ -1299,8 +1301,9 @@ static void nfs_commitdata_release(void *data)
put_nfs_open_context(wdata->args.context);
nfs_commit_free(wdata);
}
+EXPORT_SYMBOL_GPL(nfs_commitdata_release);
-static int nfs_initiate_commit(struct nfs_write_data *data, struct rpc_clnt *clnt,
+int nfs_initiate_commit(struct nfs_write_data *data, struct rpc_clnt *clnt,
const struct rpc_call_ops *call_ops,
int how)
{
@@ -1334,11 +1337,12 @@ static int nfs_initiate_commit(struct nfs_write_data *data, struct rpc_clnt *cln
rpc_put_task(task);
return 0;
}
+EXPORT_SYMBOL_GPL(nfs_initiate_commit);
/*
* Set up the argument/result storage required for the RPC call.
*/
-static void nfs_init_commit(struct nfs_write_data *data,
+void nfs_init_commit(struct nfs_write_data *data,
struct list_head *head,
struct pnfs_layout_segment *lseg)
{
@@ -1365,8 +1369,9 @@ static void nfs_init_commit(struct nfs_write_data *data,
data->res.verf = &data->verf;
nfs_fattr_init(&data->fattr);
}
+EXPORT_SYMBOL_GPL(nfs_init_commit);
-static void nfs_retry_commit(struct list_head *page_list,
+void nfs_retry_commit(struct list_head *page_list,
struct pnfs_layout_segment *lseg)
{
struct nfs_page *req;
@@ -1381,6 +1386,7 @@ static void nfs_retry_commit(struct list_head *page_list,
nfs_clear_page_tag_locked(req);
}
}
+EXPORT_SYMBOL_GPL(nfs_retry_commit);
/*
* Commit dirty pages
@@ -1419,7 +1425,7 @@ static void nfs_commit_done(struct rpc_task *task, void *calldata)
return;
}
-static void nfs_commit_release_pages(struct nfs_write_data *data)
+void nfs_commit_release_pages(struct nfs_write_data *data)
{
struct nfs_page *req;
int status = data->task.tk_status;
@@ -1456,6 +1462,7 @@ static void nfs_commit_release_pages(struct nfs_write_data *data)
nfs_clear_page_tag_locked(req);
}
}
+EXPORT_SYMBOL_GPL(nfs_commit_release_pages);
static void nfs_commit_release(void *calldata)
{
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index eddda6ce7c42..807e07c86b26 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -198,6 +198,7 @@ struct nfs_inode {
/* pNFS layout information */
struct pnfs_layout_hdr *layout;
+ atomic_t commits_outstanding;
#endif /* CONFIG_NFS_V4*/
#ifdef CONFIG_NFS_FSCACHE
struct fscache_cookie *fscache;
diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h
index 2c2c67d2eb42..ac0c0e51786e 100644
--- a/include/linux/nfs_xdr.h
+++ b/include/linux/nfs_xdr.h
@@ -1040,6 +1040,7 @@ struct nfs_write_data {
struct nfs_writeres res; /* result struct */
struct pnfs_layout_segment *lseg;
struct nfs_client *ds_clp; /* pNFS data server */
+ int ds_commit_index;
const struct rpc_call_ops *mds_ops;
int (*write_done_cb) (struct rpc_task *task, struct nfs_write_data *data);
#ifdef CONFIG_NFS_V4
--
cgit v1.2.3
From 863a3c6c686d5773f7192a4818769e15db12ce08 Mon Sep 17 00:00:00 2001
From: Andy Adamson <andros@netapp.com>
Date: Wed, 23 Mar 2011 13:27:54 +0000
Subject: NFSv4.1: layoutcommit
The filelayout driver sends LAYOUTCOMMIT only when COMMIT goes to
the data server (as opposed to the MDS) and the data server WRITE
is not NFS_FILE_SYNC.
Only whole file layout support means that there is only one IOMODE_RW layout
segment.
Signed-off-by: Andy Adamson <andros@netapp.com>
Signed-off-by: Alexandros Batsakis <batsakis@netapp.com>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Signed-off-by: Dean Hildebrand <dhildeb@us.ibm.com>
Signed-off-by: Fred Isaman <iisaman@citi.umich.edu>
Signed-off-by: Mingyang Guo <guomingyang@nrchpc.ac.cn>
Signed-off-by: Tao Guo <guotao@nrchpc.ac.cn>
Signed-off-by: Zhang Jingwang <zhangjingwang@nrchpc.ac.cn>
Tested-by: Boaz Harrosh <bharrosh@panasas.com>
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/file.c | 3 ++
fs/nfs/nfs4_fs.h | 2 +
fs/nfs/nfs4filelayout.c | 18 +++++++
fs/nfs/nfs4proc.c | 94 +++++++++++++++++++++++++++++++++++
fs/nfs/nfs4xdr.c | 129 ++++++++++++++++++++++++++++++++++++++++++++++++
fs/nfs/pnfs.c | 94 +++++++++++++++++++++++++++++++++++
fs/nfs/pnfs.h | 9 +++-
fs/nfs/write.c | 15 +++++-
include/linux/nfs4.h | 1 +
include/linux/nfs_fs.h | 1 +
include/linux/nfs_xdr.h | 23 +++++++++
11 files changed, 387 insertions(+), 2 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index d85a534b15cd..85cb95de5df5 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -326,6 +326,9 @@ nfs_file_fsync(struct file *file, int datasync)
ret = xchg(&ctx->error, 0);
if (!ret && status < 0)
ret = status;
+ if (!ret && !datasync)
+ /* application has asked for meta-data sync */
+ ret = pnfs_layoutcommit_inode(inode, 1);
return ret;
}
diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
index c64be1cff080..1e612d159b71 100644
--- a/fs/nfs/nfs4_fs.h
+++ b/fs/nfs/nfs4_fs.h
@@ -262,6 +262,8 @@ extern int nfs4_proc_destroy_session(struct nfs4_session *);
extern int nfs4_init_session(struct nfs_server *server);
extern int nfs4_proc_get_lease_time(struct nfs_client *clp,
struct nfs_fsinfo *fsinfo);
+extern int nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data,
+ int sync);
static inline bool
is_ds_only_client(struct nfs_client *clp)
diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c
index 97e75a22af72..fc1a0e9c1270 100644
--- a/fs/nfs/nfs4filelayout.c
+++ b/fs/nfs/nfs4filelayout.c
@@ -153,6 +153,23 @@ static int filelayout_read_done_cb(struct rpc_task *task,
return 0;
}
+/*
+ * We reference the rpc_cred of the first WRITE that triggers the need for
+ * a LAYOUTCOMMIT, and use it to send the layoutcommit compound.
+ * rfc5661 is not clear about which credential should be used.
+ */
+static void
+filelayout_set_layoutcommit(struct nfs_write_data *wdata)
+{
+ if (FILELAYOUT_LSEG(wdata->lseg)->commit_through_mds ||
+ wdata->res.verf->committed == NFS_FILE_SYNC)
+ return;
+
+ pnfs_set_layoutcommit(wdata);
+ dprintk("%s ionde %lu pls_end_pos %lu\n", __func__, wdata->inode->i_ino,
+ (unsigned long) wdata->lseg->pls_end_pos);
+}
+
/*
* Call ops for the async read/write cases
* In the case of dense layouts, the offset needs to be reset to its
@@ -210,6 +227,7 @@ static int filelayout_write_done_cb(struct rpc_task *task,
return -EAGAIN;
}
+ filelayout_set_layoutcommit(data);
return 0;
}
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 5d61cccc8d4d..6f2f40239d10 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -5616,6 +5616,100 @@ int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
}
EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
+static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
+{
+ struct nfs4_layoutcommit_data *data = calldata;
+ struct nfs_server *server = NFS_SERVER(data->args.inode);
+
+ if (nfs4_setup_sequence(server, &data->args.seq_args,
+ &data->res.seq_res, 1, task))
+ return;
+ rpc_call_start(task);
+}
+
+static void
+nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
+{
+ struct nfs4_layoutcommit_data *data = calldata;
+ struct nfs_server *server = NFS_SERVER(data->args.inode);
+
+ if (!nfs4_sequence_done(task, &data->res.seq_res))
+ return;
+
+ switch (task->tk_status) { /* Just ignore these failures */
+ case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
+ case NFS4ERR_BADIOMODE: /* no IOMODE_RW layout for range */
+ case NFS4ERR_BADLAYOUT: /* no layout */
+ case NFS4ERR_GRACE: /* loca_recalim always false */
+ task->tk_status = 0;
+ }
+
+ if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
+ nfs_restart_rpc(task, server->nfs_client);
+ return;
+ }
+
+ if (task->tk_status == 0)
+ nfs_post_op_update_inode_force_wcc(data->args.inode,
+ data->res.fattr);
+}
+
+static void nfs4_layoutcommit_release(void *calldata)
+{
+ struct nfs4_layoutcommit_data *data = calldata;
+
+ /* Matched by references in pnfs_set_layoutcommit */
+ put_lseg(data->lseg);
+ put_rpccred(data->cred);
+ kfree(data);
+}
+
+static const struct rpc_call_ops nfs4_layoutcommit_ops = {
+ .rpc_call_prepare = nfs4_layoutcommit_prepare,
+ .rpc_call_done = nfs4_layoutcommit_done,
+ .rpc_release = nfs4_layoutcommit_release,
+};
+
+int
+nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, int sync)
+{
+ struct rpc_message msg = {
+ .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
+ .rpc_argp = &data->args,
+ .rpc_resp = &data->res,
+ .rpc_cred = data->cred,
+ };
+ struct rpc_task_setup task_setup_data = {
+ .task = &data->task,
+ .rpc_client = NFS_CLIENT(data->args.inode),
+ .rpc_message = &msg,
+ .callback_ops = &nfs4_layoutcommit_ops,
+ .callback_data = data,
+ .flags = RPC_TASK_ASYNC,
+ };
+ struct rpc_task *task;
+ int status = 0;
+
+ dprintk("NFS: %4d initiating layoutcommit call. sync %d "
+ "lbw: %llu inode %lu\n",
+ data->task.tk_pid, sync,
+ data->args.lastbytewritten,
+ data->args.inode->i_ino);
+
+ task = rpc_run_task(&task_setup_data);
+ if (IS_ERR(task))
+ return PTR_ERR(task);
+ if (!sync)
+ goto out;
+ status = nfs4_wait_for_completion_rpc_task(task);
+ if (status != 0)
+ goto out;
+ status = task->tk_status;
+out:
+ dprintk("%s: status %d\n", __func__, status);
+ rpc_put_task(task);
+ return status;
+}
#endif /* CONFIG_NFS_V4_1 */
struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
index 07cdf925c524..207d399c8dee 100644
--- a/fs/nfs/nfs4xdr.c
+++ b/fs/nfs/nfs4xdr.c
@@ -324,6 +324,18 @@ static int nfs4_stat_to_errno(int);
#define decode_layoutget_maxsz (op_decode_hdr_maxsz + 8 + \
decode_stateid_maxsz + \
XDR_QUADLEN(PNFS_LAYOUT_MAXSIZE))
+#define encode_layoutcommit_maxsz (op_encode_hdr_maxsz + \
+ 2 /* offset */ + \
+ 2 /* length */ + \
+ 1 /* reclaim */ + \
+ encode_stateid_maxsz + \
+ 1 /* new offset (true) */ + \
+ 2 /* last byte written */ + \
+ 1 /* nt_timechanged (false) */ + \
+ 1 /* layoutupdate4 layout type */ + \
+ 1 /* NULL filelayout layoutupdate4 payload */)
+#define decode_layoutcommit_maxsz (op_decode_hdr_maxsz + 3)
+
#else /* CONFIG_NFS_V4_1 */
#define encode_sequence_maxsz 0
#define decode_sequence_maxsz 0
@@ -727,6 +739,17 @@ static int nfs4_stat_to_errno(int);
decode_sequence_maxsz + \
decode_putfh_maxsz + \
decode_layoutget_maxsz)
+#define NFS4_enc_layoutcommit_sz (compound_encode_hdr_maxsz + \
+ encode_sequence_maxsz +\
+ encode_putfh_maxsz + \
+ encode_layoutcommit_maxsz + \
+ encode_getattr_maxsz)
+#define NFS4_dec_layoutcommit_sz (compound_decode_hdr_maxsz + \
+ decode_sequence_maxsz + \
+ decode_putfh_maxsz + \
+ decode_layoutcommit_maxsz + \
+ decode_getattr_maxsz)
+
const u32 nfs41_maxwrite_overhead = ((RPC_MAX_HEADER_WITH_AUTH +
compound_encode_hdr_maxsz +
@@ -1816,6 +1839,34 @@ encode_layoutget(struct xdr_stream *xdr,
hdr->nops++;
hdr->replen += decode_layoutget_maxsz;
}
+
+static int
+encode_layoutcommit(struct xdr_stream *xdr,
+ const struct nfs4_layoutcommit_args *args,
+ struct compound_hdr *hdr)
+{
+ __be32 *p;
+
+ dprintk("%s: lbw: %llu type: %d\n", __func__, args->lastbytewritten,
+ NFS_SERVER(args->inode)->pnfs_curr_ld->id);
+
+ p = reserve_space(xdr, 48 + NFS4_STATEID_SIZE);
+ *p++ = cpu_to_be32(OP_LAYOUTCOMMIT);
+ /* Only whole file layouts */
+ p = xdr_encode_hyper(p, 0); /* offset */
+ p = xdr_encode_hyper(p, NFS4_MAX_UINT64); /* length */
+ *p++ = cpu_to_be32(0); /* reclaim */
+ p = xdr_encode_opaque_fixed(p, args->stateid.data, NFS4_STATEID_SIZE);
+ *p++ = cpu_to_be32(1); /* newoffset = TRUE */
+ p = xdr_encode_hyper(p, args->lastbytewritten);
+ *p++ = cpu_to_be32(0); /* Never send time_modify_changed */
+ *p++ = cpu_to_be32(NFS_SERVER(args->inode)->pnfs_curr_ld->id);/* type */
+ *p++ = cpu_to_be32(0); /* no file layout payload */
+
+ hdr->nops++;
+ hdr->replen += decode_layoutcommit_maxsz;
+ return 0;
+}
#endif /* CONFIG_NFS_V4_1 */
/*
@@ -2607,6 +2658,26 @@ static void nfs4_xdr_enc_layoutget(struct rpc_rqst *req,
encode_layoutget(xdr, args, &hdr);
encode_nops(&hdr);
}
+
+/*
+ * Encode LAYOUTCOMMIT request
+ */
+static int nfs4_xdr_enc_layoutcommit(struct rpc_rqst *req,
+ struct xdr_stream *xdr,
+ struct nfs4_layoutcommit_args *args)
+{
+ struct compound_hdr hdr = {
+ .minorversion = nfs4_xdr_minorversion(&args->seq_args),
+ };
+
+ encode_compound_hdr(xdr, req, &hdr);
+ encode_sequence(xdr, &args->seq_args, &hdr);
+ encode_putfh(xdr, NFS_FH(args->inode), &hdr);
+ encode_layoutcommit(xdr, args, &hdr);
+ encode_getfattr(xdr, args->bitmask, &hdr);
+ encode_nops(&hdr);
+ return 0;
+}
#endif /* CONFIG_NFS_V4_1 */
static void print_overflow_msg(const char *func, const struct xdr_stream *xdr)
@@ -5007,6 +5078,35 @@ out_overflow:
print_overflow_msg(__func__, xdr);
return -EIO;
}
+
+static int decode_layoutcommit(struct xdr_stream *xdr,
+ struct rpc_rqst *req,
+ struct nfs4_layoutcommit_res *res)
+{
+ __be32 *p;
+ __u32 sizechanged;
+ int status;
+
+ status = decode_op_hdr(xdr, OP_LAYOUTCOMMIT);
+ if (status)
+ return status;
+
+ p = xdr_inline_decode(xdr, 4);
+ if (unlikely(!p))
+ goto out_overflow;
+ sizechanged = be32_to_cpup(p);
+
+ if (sizechanged) {
+ /* throw away new size */
+ p = xdr_inline_decode(xdr, 8);
+ if (unlikely(!p))
+ goto out_overflow;
+ }
+ return 0;
+out_overflow:
+ print_overflow_msg(__func__, xdr);
+ return -EIO;
+}
#endif /* CONFIG_NFS_V4_1 */
/*
@@ -6068,6 +6168,34 @@ static int nfs4_xdr_dec_layoutget(struct rpc_rqst *rqstp,
out:
return status;
}
+
+/*
+ * Decode LAYOUTCOMMIT response
+ */
+static int nfs4_xdr_dec_layoutcommit(struct rpc_rqst *rqstp,
+ struct xdr_stream *xdr,
+ struct nfs4_layoutcommit_res *res)
+{
+ struct compound_hdr hdr;
+ int status;
+
+ status = decode_compound_hdr(xdr, &hdr);
+ if (status)
+ goto out;
+ status = decode_sequence(xdr, &res->seq_res, rqstp);
+ if (status)
+ goto out;
+ status = decode_putfh(xdr);
+ if (status)
+ goto out;
+ status = decode_layoutcommit(xdr, rqstp, res);
+ if (status)
+ goto out;
+ decode_getfattr(xdr, res->fattr, res->server,
+ !RPC_IS_ASYNC(rqstp->rq_task));
+out:
+ return status;
+}
#endif /* CONFIG_NFS_V4_1 */
/**
@@ -6269,6 +6397,7 @@ struct rpc_procinfo nfs4_procedures[] = {
PROC(RECLAIM_COMPLETE, enc_reclaim_complete, dec_reclaim_complete),
PROC(GETDEVICEINFO, enc_getdeviceinfo, dec_getdeviceinfo),
PROC(LAYOUTGET, enc_layoutget, dec_layoutget),
+ PROC(LAYOUTCOMMIT, enc_layoutcommit, dec_layoutcommit),
#endif /* CONFIG_NFS_V4_1 */
};
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index c67565965f2a..2a08ca0dddc1 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -946,3 +946,97 @@ pnfs_try_to_read_data(struct nfs_read_data *rdata,
dprintk("%s End (trypnfs:%d)\n", __func__, trypnfs);
return trypnfs;
}
+
+/*
+ * Currently there is only one (whole file) write lseg.
+ */
+static struct pnfs_layout_segment *pnfs_list_write_lseg(struct inode *inode)
+{
+ struct pnfs_layout_segment *lseg, *rv = NULL;
+
+ list_for_each_entry(lseg, &NFS_I(inode)->layout->plh_segs, pls_list)
+ if (lseg->pls_range.iomode == IOMODE_RW)
+ rv = lseg;
+ return rv;
+}
+
+void
+pnfs_set_layoutcommit(struct nfs_write_data *wdata)
+{
+ struct nfs_inode *nfsi = NFS_I(wdata->inode);
+ loff_t end_pos = wdata->args.offset + wdata->res.count;
+
+ spin_lock(&nfsi->vfs_inode.i_lock);
+ if (!test_and_set_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) {
+ /* references matched in nfs4_layoutcommit_release */
+ get_lseg(wdata->lseg);
+ wdata->lseg->pls_lc_cred =
+ get_rpccred(wdata->args.context->state->owner->so_cred);
+ mark_inode_dirty_sync(wdata->inode);
+ dprintk("%s: Set layoutcommit for inode %lu ",
+ __func__, wdata->inode->i_ino);
+ }
+ if (end_pos > wdata->lseg->pls_end_pos)
+ wdata->lseg->pls_end_pos = end_pos;
+ spin_unlock(&nfsi->vfs_inode.i_lock);
+}
+EXPORT_SYMBOL_GPL(pnfs_set_layoutcommit);
+
+int
+pnfs_layoutcommit_inode(struct inode *inode, int sync)
+{
+ struct nfs4_layoutcommit_data *data;
+ struct nfs_inode *nfsi = NFS_I(inode);
+ struct pnfs_layout_segment *lseg;
+ struct rpc_cred *cred;
+ loff_t end_pos;
+ int status = 0;
+
+ dprintk("--> %s inode %lu\n", __func__, inode->i_ino);
+
+ /* Note kzalloc ensures data->res.seq_res.sr_slot == NULL */
+ data = kzalloc(sizeof(*data), GFP_NOFS);
+ spin_lock(&inode->i_lock);
+
+ if (!test_and_clear_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) {
+ spin_unlock(&inode->i_lock);
+ kfree(data);
+ goto out;
+ }
+ /*
+ * Currently only one (whole file) write lseg which is referenced
+ * in pnfs_set_layoutcommit and will be found.
+ */
+ lseg = pnfs_list_write_lseg(inode);
+
+ end_pos = lseg->pls_end_pos;
+ cred = lseg->pls_lc_cred;
+ lseg->pls_end_pos = 0;
+ lseg->pls_lc_cred = NULL;
+
+ if (!data) {
+ put_lseg(lseg);
+ spin_unlock(&inode->i_lock);
+ put_rpccred(cred);
+ status = -ENOMEM;
+ goto out;
+ } else {
+ memcpy(&data->args.stateid.data, nfsi->layout->plh_stateid.data,
+ sizeof(nfsi->layout->plh_stateid.data));
+ }
+ spin_unlock(&inode->i_lock);
+
+ data->args.inode = inode;
+ data->lseg = lseg;
+ data->cred = cred;
+ nfs_fattr_init(&data->fattr);
+ data->args.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
+ data->res.fattr = &data->fattr;
+ data->args.lastbytewritten = end_pos - 1;
+ data->res.server = NFS_SERVER(inode);
+
+ status = nfs4_proc_layoutcommit(data, sync);
+out:
+ dprintk("<-- %s status %d\n", __func__, status);
+ return status;
+}
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index 5370f1b9aa43..0806c77862b6 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -43,6 +43,8 @@ struct pnfs_layout_segment {
atomic_t pls_refcount;
unsigned long pls_flags;
struct pnfs_layout_hdr *pls_layout;
+ struct rpc_cred *pls_lc_cred; /* LAYOUTCOMMIT credential */
+ loff_t pls_end_pos; /* LAYOUTCOMMIT write end */
};
enum pnfs_try_status {
@@ -152,7 +154,8 @@ bool pnfs_roc(struct inode *ino);
void pnfs_roc_release(struct inode *ino);
void pnfs_roc_set_barrier(struct inode *ino, u32 barrier);
bool pnfs_roc_drain(struct inode *ino, u32 *barrier);
-
+void pnfs_set_layoutcommit(struct nfs_write_data *wdata);
+int pnfs_layoutcommit_inode(struct inode *inode, int sync);
static inline int lo_fail_bit(u32 iomode)
{
@@ -325,6 +328,10 @@ static inline void pnfs_clear_request_commit(struct nfs_page *req)
{
}
+static inline int pnfs_layoutcommit_inode(struct inode *inode, int sync)
+{
+ return 0;
+}
#endif /* CONFIG_NFS_V4_1 */
#endif /* FS_NFS_PNFS_H */
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index e7aeda0663c5..a03c11f9081e 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -1562,7 +1562,20 @@ static int nfs_commit_unstable_pages(struct inode *inode, struct writeback_contr
int nfs_write_inode(struct inode *inode, struct writeback_control *wbc)
{
- return nfs_commit_unstable_pages(inode, wbc);
+ int ret;
+
+ ret = nfs_commit_unstable_pages(inode, wbc);
+ if (ret >= 0 && test_bit(NFS_INO_LAYOUTCOMMIT, &NFS_I(inode)->flags)) {
+ int status, sync = wbc->sync_mode;
+
+ if (wbc->nonblocking || wbc->for_background)
+ sync = 0;
+
+ status = pnfs_layoutcommit_inode(inode, sync);
+ if (status < 0)
+ return status;
+ }
+ return ret;
}
/*
diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h
index 134716e5e350..bf22cfaaeee5 100644
--- a/include/linux/nfs4.h
+++ b/include/linux/nfs4.h
@@ -560,6 +560,7 @@ enum {
NFSPROC4_CLNT_RECLAIM_COMPLETE,
NFSPROC4_CLNT_LAYOUTGET,
NFSPROC4_CLNT_GETDEVICEINFO,
+ NFSPROC4_CLNT_LAYOUTCOMMIT,
};
/* nfs41 types */
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index 807e07c86b26..1b93b9c60e55 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -228,6 +228,7 @@ struct nfs_inode {
#define NFS_INO_FSCACHE_LOCK (6) /* FS-Cache cookie management lock */
#define NFS_INO_COMMIT (7) /* inode is committing unstable writes */
#define NFS_INO_PNFS_COMMIT (8) /* use pnfs code for commit */
+#define NFS_INO_LAYOUTCOMMIT (9) /* layoutcommit required */
static inline struct nfs_inode *NFS_I(const struct inode *inode)
{
diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h
index ac0c0e51786e..84f3585c5728 100644
--- a/include/linux/nfs_xdr.h
+++ b/include/linux/nfs_xdr.h
@@ -236,6 +236,29 @@ struct nfs4_getdeviceinfo_res {
struct nfs4_sequence_res seq_res;
};
+struct nfs4_layoutcommit_args {
+ nfs4_stateid stateid;
+ __u64 lastbytewritten;
+ struct inode *inode;
+ const u32 *bitmask;
+ struct nfs4_sequence_args seq_args;
+};
+
+struct nfs4_layoutcommit_res {
+ struct nfs_fattr *fattr;
+ const struct nfs_server *server;
+ struct nfs4_sequence_res seq_res;
+};
+
+struct nfs4_layoutcommit_data {
+ struct rpc_task task;
+ struct nfs_fattr fattr;
+ struct pnfs_layout_segment *lseg;
+ struct rpc_cred *cred;
+ struct nfs4_layoutcommit_args args;
+ struct nfs4_layoutcommit_res res;
+};
+
/*
* Arguments to the open call.
*/
--
cgit v1.2.3
From cccb4d063b263ac0713ab27d98460fda3b4f83ff Mon Sep 17 00:00:00 2001
From: Fred Isaman <iisaman@netapp.com>
Date: Wed, 23 Mar 2011 13:27:55 +0000
Subject: NFSv4.1 remove temp code that prevented ds commits
Now that all the infrastructure is in place, we will do the
right thing if we remove this special casing.
Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/nfs4filelayout.c | 4 ----
1 file changed, 4 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c
index fc1a0e9c1270..ffb54a082f3a 100644
--- a/fs/nfs/nfs4filelayout.c
+++ b/fs/nfs/nfs4filelayout.c
@@ -385,10 +385,6 @@ filelayout_write_pagelist(struct nfs_write_data *data, int sync)
data->inode->i_ino, sync, (size_t) data->args.count, offset,
ntohl(ds->ds_ip_addr), ntohs(ds->ds_port));
- /* We can't handle commit to ds yet */
- if (!FILELAYOUT_LSEG(lseg)->commit_through_mds)
- data->args.stable = NFS_FILE_SYNC;
-
data->write_done_cb = filelayout_write_done_cb;
data->ds_clp = ds->ds_clp;
fh = nfs4_fl_select_ds_fh(lseg, j);
--
cgit v1.2.3
From de4b15c7e9fe9fdd1eb6333be0a08c822db1ca0e Mon Sep 17 00:00:00 2001
From: Andy Adamson <andros@netapp.com>
Date: Sat, 12 Mar 2011 02:58:09 -0500
Subject: NFSv4.1 pnfs_layoutcommit_inode fixes
Test NFS_INO_LAYOUTCOMMIT before kzalloc
Mark inode dirty to retry LAYOUTCOMMIT on kzalloc failure.
Add comments.
Signed-off-by: Andy Adamson <andros@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/pnfs.c | 30 +++++++++++++++++++-----------
1 file changed, 19 insertions(+), 11 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 2a08ca0dddc1..ac7112574790 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -982,6 +982,14 @@ pnfs_set_layoutcommit(struct nfs_write_data *wdata)
}
EXPORT_SYMBOL_GPL(pnfs_set_layoutcommit);
+/*
+ * For the LAYOUT4_NFSV4_1_FILES layout type, NFS_DATA_SYNC WRITEs and
+ * NFS_UNSTABLE WRITEs with a COMMIT to data servers must store enough
+ * data to disk to allow the server to recover the data if it crashes.
+ * LAYOUTCOMMIT is only needed when the NFL4_UFLG_COMMIT_THRU_MDS flag
+ * is off, and a COMMIT is sent to a data server, or
+ * if WRITEs to a data server return NFS_DATA_SYNC.
+ */
int
pnfs_layoutcommit_inode(struct inode *inode, int sync)
{
@@ -994,10 +1002,18 @@ pnfs_layoutcommit_inode(struct inode *inode, int sync)
dprintk("--> %s inode %lu\n", __func__, inode->i_ino);
+ if (!test_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags))
+ return 0;
+
/* Note kzalloc ensures data->res.seq_res.sr_slot == NULL */
data = kzalloc(sizeof(*data), GFP_NOFS);
- spin_lock(&inode->i_lock);
+ if (!data) {
+ mark_inode_dirty_sync(inode);
+ status = -ENOMEM;
+ goto out;
+ }
+ spin_lock(&inode->i_lock);
if (!test_and_clear_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) {
spin_unlock(&inode->i_lock);
kfree(data);
@@ -1014,16 +1030,8 @@ pnfs_layoutcommit_inode(struct inode *inode, int sync)
lseg->pls_end_pos = 0;
lseg->pls_lc_cred = NULL;
- if (!data) {
- put_lseg(lseg);
- spin_unlock(&inode->i_lock);
- put_rpccred(cred);
- status = -ENOMEM;
- goto out;
- } else {
- memcpy(&data->args.stateid.data, nfsi->layout->plh_stateid.data,
- sizeof(nfsi->layout->plh_stateid.data));
- }
+ memcpy(&data->args.stateid.data, nfsi->layout->plh_stateid.data,
+ sizeof(nfsi->layout->plh_stateid.data));
spin_unlock(&inode->i_lock);
data->args.inode = inode;
--
cgit v1.2.3
From ef31153786bc1e4304e6b9422cc8b9efef455611 Mon Sep 17 00:00:00 2001
From: Andy Adamson <andros@netapp.com>
Date: Sat, 12 Mar 2011 02:58:10 -0500
Subject: NFSv4.1 convert layoutcommit sync to boolean
Signed-off-by: Andy Adamson <andros@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/file.c | 2 +-
fs/nfs/nfs4_fs.h | 2 +-
fs/nfs/nfs4proc.c | 4 ++--
fs/nfs/pnfs.c | 2 +-
fs/nfs/pnfs.h | 4 ++--
fs/nfs/write.c | 8 +++++---
6 files changed, 12 insertions(+), 10 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index 85cb95de5df5..3ac5bd695e5e 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -328,7 +328,7 @@ nfs_file_fsync(struct file *file, int datasync)
ret = status;
if (!ret && !datasync)
/* application has asked for meta-data sync */
- ret = pnfs_layoutcommit_inode(inode, 1);
+ ret = pnfs_layoutcommit_inode(inode, true);
return ret;
}
diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
index 1e612d159b71..4414fd7e7d5e 100644
--- a/fs/nfs/nfs4_fs.h
+++ b/fs/nfs/nfs4_fs.h
@@ -263,7 +263,7 @@ extern int nfs4_init_session(struct nfs_server *server);
extern int nfs4_proc_get_lease_time(struct nfs_client *clp,
struct nfs_fsinfo *fsinfo);
extern int nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data,
- int sync);
+ bool sync);
static inline bool
is_ds_only_client(struct nfs_client *clp)
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 6f2f40239d10..43045fa44710 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -5671,7 +5671,7 @@ static const struct rpc_call_ops nfs4_layoutcommit_ops = {
};
int
-nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, int sync)
+nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
{
struct rpc_message msg = {
.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
@@ -5699,7 +5699,7 @@ nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, int sync)
task = rpc_run_task(&task_setup_data);
if (IS_ERR(task))
return PTR_ERR(task);
- if (!sync)
+ if (sync == false)
goto out;
status = nfs4_wait_for_completion_rpc_task(task);
if (status != 0)
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index ac7112574790..22c2ddbef420 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -991,7 +991,7 @@ EXPORT_SYMBOL_GPL(pnfs_set_layoutcommit);
* if WRITEs to a data server return NFS_DATA_SYNC.
*/
int
-pnfs_layoutcommit_inode(struct inode *inode, int sync)
+pnfs_layoutcommit_inode(struct inode *inode, bool sync)
{
struct nfs4_layoutcommit_data *data;
struct nfs_inode *nfsi = NFS_I(inode);
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index 0806c77862b6..33b9ae90c6f7 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -155,7 +155,7 @@ void pnfs_roc_release(struct inode *ino);
void pnfs_roc_set_barrier(struct inode *ino, u32 barrier);
bool pnfs_roc_drain(struct inode *ino, u32 *barrier);
void pnfs_set_layoutcommit(struct nfs_write_data *wdata);
-int pnfs_layoutcommit_inode(struct inode *inode, int sync);
+int pnfs_layoutcommit_inode(struct inode *inode, bool sync);
static inline int lo_fail_bit(u32 iomode)
{
@@ -328,7 +328,7 @@ static inline void pnfs_clear_request_commit(struct nfs_page *req)
{
}
-static inline int pnfs_layoutcommit_inode(struct inode *inode, int sync)
+static inline int pnfs_layoutcommit_inode(struct inode *inode, bool sync)
{
return 0;
}
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index a03c11f9081e..85d75254328e 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -1566,10 +1566,12 @@ int nfs_write_inode(struct inode *inode, struct writeback_control *wbc)
ret = nfs_commit_unstable_pages(inode, wbc);
if (ret >= 0 && test_bit(NFS_INO_LAYOUTCOMMIT, &NFS_I(inode)->flags)) {
- int status, sync = wbc->sync_mode;
+ int status;
+ bool sync = true;
- if (wbc->nonblocking || wbc->for_background)
- sync = 0;
+ if (wbc->sync_mode == WB_SYNC_NONE || wbc->nonblocking ||
+ wbc->for_background)
+ sync = false;
status = pnfs_layoutcommit_inode(inode, sync);
if (status < 0)
--
cgit v1.2.3
From 35124a0994fc02545b14b9fa3aad000b3331f1c0 Mon Sep 17 00:00:00 2001
From: Weston Andros Adamson <dros@netapp.com>
Date: Thu, 24 Mar 2011 16:48:21 -0400
Subject: Cleanup XDR parsing for LAYOUTGET, GETDEVICEINFO
changes LAYOUTGET and GETDEVICEINFO XDR parsing to:
- not use vmap, which doesn't work on incoherent archs
- use xdr_stream parsing for all xdr
Signed-off-by: Weston Andros Adamson <dros@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
fs/nfs/nfs4filelayout.c | 53 +++++++++++---
fs/nfs/nfs4filelayoutdev.c | 178 +++++++++++++++++++++++++++++++--------------
fs/nfs/nfs4proc.c | 9 +--
fs/nfs/nfs4xdr.c | 30 +++++---
fs/nfs/pnfs.c | 39 ++++++++++
fs/nfs/pnfs.h | 1 -
include/linux/nfs_xdr.h | 6 +-
7 files changed, 230 insertions(+), 86 deletions(-)
(limited to 'fs')
diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c
index ffb54a082f3a..6f8192f4cfc7 100644
--- a/fs/nfs/nfs4filelayout.c
+++ b/fs/nfs/nfs4filelayout.c
@@ -502,12 +502,33 @@ filelayout_decode_layout(struct pnfs_layout_hdr *flo,
struct nfs4_layoutget_res *lgr,
struct nfs4_deviceid *id)
{
- uint32_t *p = (uint32_t *)lgr->layout.buf;
+ struct xdr_stream stream;
+ struct xdr_buf buf = {
+ .pages = lgr->layoutp->pages,
+ .page_len = lgr->layoutp->len,
+ .buflen = lgr->layoutp->len,
+ .len = lgr->layoutp->len,
+ };
+ struct page *scratch;
+ __be32 *p;
uint32_t nfl_util;
int i;
dprintk("%s: set_layout_map Begin\n", __func__);
+ scratch = alloc_page(GFP_KERNEL);
+ if (!scratch)
+ return -ENOMEM;
+
+ xdr_init_decode(&stream, &buf, NULL);
+ xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE);
+
+ /* 20 = ufl_util (4), first_stripe_index (4), pattern_offset (8),
+ * num_fh (4) */
+ p = xdr_inline_decode(&stream, NFS4_DEVICEID4_SIZE + 20);
+ if (unlikely(!p))
+ goto out_err;
+
memcpy(id, p, sizeof(*id));
p += XDR_QUADLEN(NFS4_DEVICEID4_SIZE);
print_deviceid(id);
@@ -529,32 +550,46 @@ filelayout_decode_layout(struct pnfs_layout_hdr *flo,
__func__, nfl_util, fl->num_fh, fl->first_stripe_index,
fl->pattern_offset);
+ if (!fl->num_fh)
+ goto out_err;
+
fl->fh_array = kzalloc(fl->num_fh * sizeof(struct nfs_fh *),
GFP_KERNEL);
if (!fl->fh_array)
- return -ENOMEM;
+ goto out_err;
for (i = 0; i < fl->num_fh; i++) {
/* Do we want to use a mempool here? */
fl->fh_array[i] = kmalloc(sizeof(struct nfs_fh), GFP_KERNEL);
- if (!fl->fh_array[i]) {
- filelayout_free_fh_array(fl);
- return -ENOMEM;
- }
+ if (!fl->fh_array[i])
+ goto out_err_free;
+
+ p = xdr_inline_decode(&stream, 4);
+ if (unlikely(!p))
+ goto out_err_free;
fl->fh_array[i]->size = be32_to_cpup(p++);
if (sizeof(struct nfs_fh) < fl->fh_array[i]->size) {
printk(KERN_ERR "Too big fh %d received %d\n",
i, fl->fh_array[i]->size);
- filelayout_free_fh_array(fl);
- return -EIO;
+ goto out_err_free;
}
+
+ p = xdr_inline_decode(&stream, fl->fh_array[i]->size);
+ if (unlikely(!p))
+ goto out_err_free;
memcpy(fl->fh_array[i]->data, p, fl->fh_array[i]->size);
- p += XDR_QUADLEN(fl->fh_array[i]->size);
dprintk("DEBUG: %s: fh len %d\n", __func__,
fl->fh_array[i]->size);
}
+ __free_page(scratch);
return 0;
+
+out_err_free:
+ filelayout_free_fh_array(fl);
+out_err:
+ __free_page(scratch);
+ return -EIO;
}
static void
diff --git a/fs/nfs/nfs4filelayoutdev.c b/fs/nfs/nfs4filelayoutdev.c
index 68143c162e3b..de5350f2b249 100644
--- a/fs/nfs/nfs4filelayoutdev.c
+++ b/fs/nfs/nfs4filelayoutdev.c
@@ -261,7 +261,7 @@ out:
* Currently only support ipv4, and one multi-path address.
*/
static struct nfs4_pnfs_ds *
-decode_and_add_ds(__be32 **pp, struct inode *inode)
+decode_and_add_ds(struct xdr_stream *streamp, struct inode *inode)
{
struct nfs4_pnfs_ds *ds = NULL;
char *buf;
@@ -269,25 +269,34 @@ decode_and_add_ds(__be32 **pp, struct inode *inode)
u32 ip_addr, port;
int nlen, rlen, i;
int tmp[2];
- __be32 *r_netid, *r_addr, *p = *pp;
+ __be32 *p;
/* r_netid */
+ p = xdr_inline_decode(streamp, 4);
+ if (unlikely(!p))
+ goto out_err;
nlen = be32_to_cpup(p++);
- r_netid = p;
- p += XDR_QUADLEN(nlen);
- /* r_addr */
- rlen = be32_to_cpup(p++);
- r_addr = p;
- p += XDR_QUADLEN(rlen);
- *pp = p;
+ p = xdr_inline_decode(streamp, nlen);
+ if (unlikely(!p))
+ goto out_err;
/* Check that netid is "tcp" */
- if (nlen != 3 || memcmp((char *)r_netid, "tcp", 3)) {
+ if (nlen != 3 || memcmp((char *)p, "tcp", 3)) {
dprintk("%s: ERROR: non ipv4 TCP r_netid\n", __func__);
goto out_err;
}
+ /* r_addr */
+ p = xdr_inline_decode(streamp, 4);
+ if (unlikely(!p))
+ goto out_err;
+ rlen = be32_to_cpup(p);
+
+ p = xdr_inline_decode(streamp, rlen);
+ if (unlikely(!p))
+ goto out_err;
+
/* ipv6 length plus port is legal */
if (rlen > INET6_ADDRSTRLEN + 8) {
dprintk("%s: Invalid address, length %d\n", __func__,
@@ -300,7 +309,7 @@ decode_and_add_ds(__be32 **pp, struct inode *inode)
goto out_err;
}
buf[rlen] = '\0';
- memcpy(buf, r_addr, rlen);
+ memcpy(buf, p, rlen);
/* replace the port dots with dashes for the in4_pton() delimiter*/
for (i = 0; i < 2; i++) {
@@ -336,90 +345,154 @@ out_err:
static struct nfs4_file_layout_dsaddr*
decode_device(struct inode *ino, struct pnfs_device *pdev)
{
- int i, dummy;
+ int i;
u32 cnt, num;
u8 *indexp;
- __be32 *p = (__be32 *)pdev->area, *indicesp;
- struct nfs4_file_layout_dsaddr *dsaddr;
+ __be32 *p;
+ u8 *stripe_indices;
+ u8 max_stripe_index;
+ struct nfs4_file_layout_dsaddr *dsaddr = NULL;
+ struct xdr_stream stream;
+ struct xdr_buf buf = {
+ .pages = pdev->pages,
+ .page_len = pdev->pglen,
+ .buflen = pdev->pglen,
+ .len = pdev->pglen,
+ };
+ struct page *scratch;
+
+ /* set up xdr stream */
+ scratch = alloc_page(GFP_KERNEL);
+ if (!scratch)
+ goto out_err;
+
+ xdr_init_decode(&stream, &buf, NULL);
+ xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE);
/* Get the stripe count (number of stripe index) */
- cnt = be32_to_cpup(p++);
+ p = xdr_inline_decode(&stream, 4);
+ if (unlikely(!p))
+ goto out_err_free_scratch;
+
+ cnt = be32_to_cpup(p);
dprintk("%s stripe count %d\n", __func__, cnt);
if (cnt > NFS4_PNFS_MAX_STRIPE_CNT) {
printk(KERN_WARNING "%s: stripe count %d greater than "
"supported maximum %d\n", __func__,
cnt, NFS4_PNFS_MAX_STRIPE_CNT);
- goto out_err;
+ goto out_err_free_scratch;
+ }
+
+ /* read stripe indices */
+ stripe_indices = kcalloc(cnt, sizeof(u8), GFP_KERNEL);
+ if (!stripe_indices)
+ goto out_err_free_scratch;
+
+ p = xdr_inline_decode(&stream, cnt << 2);
+ if (unlikely(!p))
+ goto out_err_free_stripe_indices;
+
+ indexp = &stripe_indices[0];
+ max_stripe_index = 0;
+ for (i = 0; i < cnt; i++) {
+ *indexp = be32_to_cpup(p++);
+ max_stripe_index = max(max_stripe_index, *indexp);
+ indexp++;
}
/* Check the multipath list count */
- indicesp = p;
- p += XDR_QUADLEN(cnt << 2);
- num = be32_to_cpup(p++);
+ p = xdr_inline_decode(&stream, 4);
+ if (unlikely(!p))
+ goto out_err_free_stripe_indices;
+
+ num = be32_to_cpup(p);
dprintk("%s ds_num %u\n", __func__, num);
if (num > NFS4_PNFS_MAX_MULTI_CNT) {
printk(KERN_WARNING "%s: multipath count %d greater than "
"supported maximum %d\n", __func__,
num, NFS4_PNFS_MAX_MULTI_CNT);
- goto out_err;
+ goto out_err_free_stripe_indices;
}
+
+ /* validate stripe indices are all < num */
+ if (max_stripe_index >= num) {
+ printk(KERN_WARNING "%s: stripe index %u >= num ds %u\n",
+ __func__, max_stripe_index, num);
+ goto out_err_free_stripe_indices;
+ }
+
dsaddr = kzalloc(sizeof(*dsaddr) +
(sizeof(struct nfs4_pnfs_ds *) * (num - 1)),
GFP_KERNEL);
if (!dsaddr)
- goto out_err;
-
- dsaddr->stripe_indices = kzalloc(sizeof(u8) * cnt, GFP_KERNEL);
- if (!dsaddr->stripe_indices)
- goto out_err_free;
+ goto out_err_free_stripe_indices;
dsaddr->stripe_count = cnt;
+ dsaddr->stripe_indices = stripe_indices;
+ stripe_indices = NULL;
dsaddr->ds_num = num;
memcpy(&dsaddr->deviceid, &pdev->dev_id, sizeof(pdev->dev_id));
- /* Go back an read stripe indices */
- p = indicesp;
- indexp = &dsaddr->stripe_indices[0];
- for (i = 0; i < dsaddr->stripe_count; i++) {
- *indexp = be32_to_cpup(p++);
- if (*indexp >= num)
- goto out_err_free;
- indexp++;
- }
- /* Skip already read multipath list count */
- p++;
-
for (i = 0; i < dsaddr->ds_num; i++) {
int j;
+ u32 mp_count;
+
+ p = xdr_inline_decode(&stream, 4);
+ if (unlikely(!p))
+ goto out_err_free_deviceid;
- dummy = be32_to_cpup(p++); /* multipath count */
- if (dummy > 1) {
+ mp_count = be32_to_cpup(p); /* multipath count */
+ if (mp_count > 1) {
printk(KERN_WARNING
"%s: Multipath count %d not supported, "
"skipping all greater than 1\n", __func__,
- dummy);
+ mp_count);
}
- for (j = 0; j < dummy; j++) {
+ for (j = 0; j < mp_count; j++) {
if (j == 0) {
- dsaddr->ds_list[i] = decode_and_add_ds(&p, ino);
+ dsaddr->ds_list[i] = decode_and_add_ds(&stream,
+ ino);
if (dsaddr->ds_list[i] == NULL)
- goto out_err_free;
+ goto out_err_free_deviceid;
} else {
u32 len;
/* skip extra multipath */
- len = be32_to_cpup(p++);
- p += XDR_QUADLEN(len);
- len = be32_to_cpup(p++);
- p += XDR_QUADLEN(len);
- continue;
+
+ /* read len, skip */
+ p = xdr_inline_decode(&stream, 4);
+ if (unlikely(!p))
+ goto out_err_free_deviceid;
+ len = be32_to_cpup(p);
+
+ p = xdr_inline_decode(&stream, len);
+ if (unlikely(!p))
+ goto out_err_free_deviceid;
+
+ /* read len, skip */
+ p = xdr_inline_decode(&stream, 4);
+ if (unlikely(!p))
+ goto out_err_free_deviceid;
+ len = be32_to_cpup(p);
+
+ p = xdr_inline_decode(&stream, len);
+ if (unlikely(!p))
+ goto out_err_free_deviceid;
}
}
}
+
+ __free_page(scratch);
return dsaddr;
-out_err_free:
+out_err_free_deviceid:
nfs4_fl_free_deviceid(dsaddr);
+ /* stripe_indicies was part of dsaddr */
+ goto out_err_free_scratch;
+out_err_free_stripe_indices:
+ kfree(stripe_indices);
+out_err_free_scratch:
+ __free_page(scratch);
out_err:
dprintk("%s ERROR: returning NULL\n", __func__);
return NULL;
@@ -498,11 +571,6 @@ get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id)
goto out_free;
}
- /* set pdev->area */
- pdev->area = vmap(pages, max_pages, VM_MAP, PAGE_KERNEL);
- if (!pdev->area)
- goto out_free;
-
memcpy(&pdev->dev_id, dev_id, sizeof(*dev_id));
pdev->layout_type = LAYOUT_NFSV4_1_FILES;
pdev->pages = pages;
@@ -521,8 +589,6 @@ get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id)
*/
dsaddr = decode_and_add_device(inode, pdev);
out_free:
- if (pdev->area != NULL)
- vunmap(pdev->area);
for (i = 0; i < max_pages; i++)
__free_page(pages[i]);
kfree(pages);
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 43045fa44710..8f071314e94b 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -5526,8 +5526,6 @@ static void nfs4_layoutget_release(void *calldata)
struct nfs4_layoutget *lgp = calldata;
dprintk("--> %s\n", __func__);
- if (lgp->res.layout.buf != NULL)
- free_page((unsigned long) lgp->res.layout.buf);
put_nfs_open_context(lgp->args.ctx);
kfree(calldata);
dprintk("<-- %s\n", __func__);
@@ -5559,12 +5557,7 @@ int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
dprintk("--> %s\n", __func__);
- lgp->res.layout.buf = (void *)__get_free_page(GFP_NOFS);
- if (lgp->res.layout.buf == NULL) {
- nfs4_layoutget_release(lgp);
- return -ENOMEM;
- }
-
+ lgp->res.layoutp = &lgp->args.layout;
lgp->res.seq_res.sr_slot = NULL;
task = rpc_run_task(&task_setup_data);
if (IS_ERR(task))
diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
index 207d399c8dee..40da65e8fa2a 100644
--- a/fs/nfs/nfs4xdr.c
+++ b/fs/nfs/nfs4xdr.c
@@ -2656,6 +2656,10 @@ static void nfs4_xdr_enc_layoutget(struct rpc_rqst *req,
encode_sequence(xdr, &args->seq_args, &hdr);
encode_putfh(xdr, NFS_FH(args->inode), &hdr);
encode_layoutget(xdr, args, &hdr);
+
+ xdr_inline_pages(&req->rq_rcv_buf, hdr.replen << 2,
+ args->layout.pages, 0, args->layout.pglen);
+
encode_nops(&hdr);
}
@@ -5022,6 +5026,9 @@ static int decode_layoutget(struct xdr_stream *xdr, struct rpc_rqst *req,
__be32 *p;
int status;
u32 layout_count;
+ struct xdr_buf *rcvbuf = &req->rq_rcv_buf;
+ struct kvec *iov = rcvbuf->head;
+ u32 hdrlen, recvd;
status = decode_op_hdr(xdr, OP_LAYOUTGET);
if (status)
@@ -5038,17 +5045,14 @@ static int decode_layoutget(struct xdr_stream *xdr, struct rpc_rqst *req,
return -EINVAL;
}
- p = xdr_inline_decode(xdr, 24);
+ p = xdr_inline_decode(xdr, 28);
if (unlikely(!p))
goto out_overflow;
p = xdr_decode_hyper(p, &res->range.offset);
p = xdr_decode_hyper(p, &res->range.length);
res->range.iomode = be32_to_cpup(p++);
res->type = be32_to_cpup(p++);
-
- status = decode_opaque_inline(xdr, &res->layout.len, (char **)&p);
- if (unlikely(status))
- return status;
+ res->layoutp->len = be32_to_cpup(p);
dprintk("%s roff:%lu rlen:%lu riomode:%d, lo_type:0x%x, lo.len:%d\n",
__func__,
@@ -5056,12 +5060,18 @@ static int decode_layoutget(struct xdr_stream *xdr, struct rpc_rqst *req,
(unsigned long)res->range.length,
res->range.iomode,
res->type,
- res->layout.len);
+ res->layoutp->len);
+
+ hdrlen = (u8 *) xdr->p - (u8 *) iov->iov_base;
+ recvd = req->rq_rcv_buf.len - hdrlen;
+ if (res->layoutp->len > recvd) {
+ dprintk("NFS: server cheating in layoutget reply: "
+ "layout len %u > recvd %u\n",
+ res->layoutp->len, recvd);
+ return -EINVAL;
+ }
- /* nfs4_proc_layoutget allocated a single page */
- if (res->layout.len > PAGE_SIZE)
- return -ENOMEM;
- memcpy(res->layout.buf, p, res->layout.len);
+ xdr_read_pages(xdr, res->layoutp->len);
if (layout_count > 1) {
/* We only handle a length one array at the moment. Any
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 22c2ddbef420..d9ab97269ce6 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -472,6 +472,9 @@ send_layoutget(struct pnfs_layout_hdr *lo,
struct nfs_server *server = NFS_SERVER(ino);
struct nfs4_layoutget *lgp;
struct pnfs_layout_segment *lseg = NULL;
+ struct page **pages = NULL;
+ int i;
+ u32 max_resp_sz, max_pages;
dprintk("--> %s\n", __func__);
@@ -479,6 +482,21 @@ send_layoutget(struct pnfs_layout_hdr *lo,
lgp = kzalloc(sizeof(*lgp), GFP_KERNEL);
if (lgp == NULL)
return NULL;
+
+ /* allocate pages for xdr post processing */
+ max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
+ max_pages = max_resp_sz >> PAGE_SHIFT;
+
+ pages = kzalloc(max_pages * sizeof(struct page *), GFP_KERNEL);
+ if (!pages)
+ goto out_err_free;
+
+ for (i = 0; i < max_pages; i++) {
+ pages[i] = alloc_page(GFP_KERNEL);
+ if (!pages[i])
+ goto out_err_free;
+ }
+
lgp->args.minlength = NFS4_MAX_UINT64;
lgp->args.maxcount = PNFS_LAYOUT_MAXSIZE;
lgp->args.range.iomode = iomode;
@@ -487,6 +505,8 @@ send_layoutget(struct pnfs_layout_hdr *lo,
lgp->args.type = server->pnfs_curr_ld->id;
lgp->args.inode = ino;
lgp->args.ctx = get_nfs_open_context(ctx);
+ lgp->args.layout.pages = pages;
+ lgp->args.layout.pglen = max_pages * PAGE_SIZE;
lgp->lsegpp = &lseg;
/* Synchronously retrieve layout information from server and
@@ -497,7 +517,26 @@ send_layoutget(struct pnfs_layout_hdr *lo,
/* remember that LAYOUTGET failed and suspend trying */
set_bit(lo_fail_bit(iomode), &lo->plh_flags);
}
+
+ /* free xdr pages */
+ for (i = 0; i < max_pages; i++)
+ __free_page(pages[i]);
+ kfree(pages);
+
return lseg;
+
+out_err_free:
+ /* free any allocated xdr pages, lgp as it's not used */
+ if (pages) {
+ for (i = 0; i < max_pages; i++) {
+ if (!pages[i])
+ break;
+ __free_page(pages[i]);
+ }
+ kfree(pages);
+ }
+ kfree(lgp);
+ return NULL;
}
bool pnfs_roc(struct inode *ino)
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index 33b9ae90c6f7..bc4827202e7a 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -109,7 +109,6 @@ struct pnfs_device {
unsigned int layout_type;
unsigned int mincount;
struct page **pages;
- void *area;
unsigned int pgbase;
unsigned int pglen;
};
diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h
index 84f3585c5728..a6e21b10f43d 100644
--- a/include/linux/nfs_xdr.h
+++ b/include/linux/nfs_xdr.h
@@ -190,8 +190,9 @@ struct nfs4_get_lease_time_res {
#define PNFS_LAYOUT_MAXSIZE 4096
struct nfs4_layoutdriver_data {
+ struct page **pages;
+ __u32 pglen;
__u32 len;
- void *buf;
};
struct pnfs_layout_range {
@@ -209,6 +210,7 @@ struct nfs4_layoutget_args {
struct nfs_open_context *ctx;
struct nfs4_sequence_args seq_args;
nfs4_stateid stateid;
+ struct nfs4_layoutdriver_data layout;
};
struct nfs4_layoutget_res {
@@ -216,8 +218,8 @@ struct nfs4_layoutget_res {
struct pnfs_layout_range range;
__u32 type;
nfs4_stateid stateid;
- struct nfs4_layoutdriver_data layout;
struct nfs4_sequence_res seq_res;
+ struct nfs4_layoutdriver_data *layoutp;
};
struct nfs4_layoutget {
--
cgit v1.2.3