From: Fred Isaman <iisaman@xxxxxxxxxxxxxx>
This adds the hooks in nfs_write_begin and nfs_write_end needed
by the block server
Signed-off-by: Fred Isaman <iisaman@xxxxxxxxxxxxxx>
[pnfs: prevent offset overflow in _pnfs_do_flush]
[pnfs: pnfs_has_layout take_ref parameter should be bool]
[pnfs: clean up put_unlock_current_layout's interface]
[pnfs: introduce lseg valid bit]
[pnfsblock: cleanup pnfs_free_fsdata]
Signed-off-by: Benny Halevy <bhalevy@xxxxxxxxxxx>
---
fs/nfs/file.c | 18 +++++++++++++-
fs/nfs/pnfs.c | 41 +++++++++++++++++++++++++++++++++
fs/nfs/pnfs.h | 71 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
3 files changed, 129 insertions(+), 1 deletions(-)
diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index 2f093ed..8c54b32 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -384,12 +384,15 @@ static int nfs_write_begin(struct file *file, struct address_space *mapping,
pgoff_t index = pos >> PAGE_CACHE_SHIFT;
struct page *page;
int once_thru = 0;
+ struct pnfs_layout_segment *lseg;
dfprintk(PAGECACHE, "NFS: write_begin(%s/%s(%ld), %u@%lld)\n",
file->f_path.dentry->d_parent->d_name.name,
file->f_path.dentry->d_name.name,
mapping->host->i_ino, len, (long long) pos);
-
+ lseg = pnfs_update_layout(mapping->host,
+ nfs_file_open_context(file),
+ pos, len, IOMODE_RW, GFP_NOFS);
start:
/*
* Prevent starvation issues if someone is doing a consistency
@@ -409,6 +412,9 @@ start:
if (ret) {
unlock_page(page);
page_cache_release(page);
+ *pagep = NULL;
+ *fsdata = NULL;
+ goto out;
} else if (!once_thru &&
nfs_want_read_modify_write(file, page, pos, len)) {
once_thru = 1;
@@ -417,6 +423,12 @@ start:
if (!ret)
goto start;
}
+ ret = pnfs_write_begin(file, page, pos, len, lseg, fsdata);
+ out:
+ if (ret) {
+ put_lseg(lseg);
+ *fsdata = NULL;
+ }
return ret;
}
@@ -426,6 +438,7 @@ static int nfs_write_end(struct file *file, struct address_space *mapping,
{
unsigned offset = pos & (PAGE_CACHE_SIZE - 1);
int status;
+ struct pnfs_layout_segment *lseg;
dfprintk(PAGECACHE, "NFS: write_end(%s/%s(%ld), %u@%lld)\n",
file->f_path.dentry->d_parent->d_name.name,
@@ -452,10 +465,13 @@ static int nfs_write_end(struct file *file, struct address_space *mapping,
zero_user_segment(page, pglen, PAGE_CACHE_SIZE);
}
+ lseg = nfs4_pull_lseg_from_fsdata(file, fsdata);
status = nfs_updatepage(file, page, offset, copied);
unlock_page(page);
page_cache_release(page);
+ pnfs_write_end_cleanup(file, fsdata);
+ put_lseg(lseg);
if (status < 0)
return status;
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 12b228a..c88a8ee 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -1138,6 +1138,41 @@ pnfs_try_to_write_data(struct nfs_write_data *wdata,
}
/*
+ * This gives the layout driver an opportunity to read in page "around"
+ * the data to be written. It returns 0 on success, otherwise an error code
+ * which will either be passed up to user, or ignored if
+ * some previous part of write succeeded.
+ * Note the range [pos, pos+len-1] is entirely within the page.
+ */
+int _pnfs_write_begin(struct inode *inode, struct page *page,
+ loff_t pos, unsigned len,
+ struct pnfs_layout_segment *lseg,
+ struct pnfs_fsdata **fsdata)
+{
+ struct pnfs_fsdata *data;
+ int status = 0;
+
+ dprintk("--> %s: pos=%llu len=%u\n",
+ __func__, (unsigned long long)pos, len);
+ data = kzalloc(sizeof(struct pnfs_fsdata), GFP_KERNEL);
+ if (!data) {
+ status = -ENOMEM;
+ goto out;
+ }
+ data->lseg = lseg; /* refcount passed into data to be managed there */
+ status = NFS_SERVER(inode)->pnfs_curr_ld->write_begin(
+ lseg, page, pos, len, data);
+ if (status) {
+ kfree(data);
+ data = NULL;
+ }
+out:
+ *fsdata = data;
+ dprintk("<-- %s: status=%d\n", __func__, status);
+ return status;
+}
+
+/*
* Called by non rpc-based layout drivers
*/
int
@@ -1250,6 +1285,12 @@ pnfs_set_layoutcommit(struct nfs_write_data *wdata)
}
EXPORT_SYMBOL_GPL(pnfs_set_layoutcommit);
+void pnfs_free_fsdata(struct pnfs_fsdata *fsdata)
+{
+ /* lseg refcounting handled directly in nfs_write_end */
+ kfree(fsdata);
+}
+
/*
* 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
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index b3481e5..5712053 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -54,6 +54,10 @@ enum pnfs_try_status {
PNFS_NOT_ATTEMPTED = 1,
};
+struct pnfs_fsdata {
+ struct pnfs_layout_segment *lseg;
+};
+
#ifdef CONFIG_NFS_V4_1
#define LAYOUT_NFSV4_1_MODULE_PREFIX "nfs-layouttype4"
@@ -106,6 +110,9 @@ struct pnfs_layoutdriver_type {
*/
enum pnfs_try_status (*read_pagelist) (struct nfs_read_data *nfs_data);
enum pnfs_try_status (*write_pagelist) (struct nfs_write_data *nfs_data, int how);
+ int (*write_begin) (struct pnfs_layout_segment *lseg, struct page *page,
+ loff_t pos, unsigned count,
+ struct pnfs_fsdata *fsdata);
void (*free_deviceid_node) (struct nfs4_deviceid_node *);
@@ -180,6 +187,7 @@ enum pnfs_try_status pnfs_try_to_write_data(struct nfs_write_data *,
enum pnfs_try_status pnfs_try_to_read_data(struct nfs_read_data *,
const struct rpc_call_ops *);
bool pnfs_generic_pg_test(struct nfs_pageio_descriptor *pgio, struct nfs_page *prev, struct nfs_page *req);
+void pnfs_free_fsdata(struct pnfs_fsdata *fsdata);
void pnfs_cleanup_layoutcommit(struct inode *,
struct nfs4_layoutcommit_data *);
int pnfs_layout_process(struct nfs4_layoutget *lgp);
@@ -193,6 +201,10 @@ void pnfs_set_layout_stateid(struct pnfs_layout_hdr *lo,
int pnfs_choose_layoutget_stateid(nfs4_stateid *dst,
struct pnfs_layout_hdr *lo,
struct nfs4_state *open_state);
+int _pnfs_write_begin(struct inode *inode, struct page *page,
+ loff_t pos, unsigned len,
+ struct pnfs_layout_segment *lseg,
+ struct pnfs_fsdata **fsdata);
int mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo,
struct list_head *tmp_list,
struct pnfs_layout_range *recall_range);
@@ -305,6 +317,32 @@ pnfs_ld_layoutret_on_setattr(struct inode *inode)
PNFS_LAYOUTRET_ON_SETATTR;
}
+static inline int pnfs_write_begin(struct file *filp, struct page *page,
+ loff_t pos, unsigned len,
+ struct pnfs_layout_segment *lseg,
+ void **fsdata)
+{
+ struct inode *inode = filp->f_dentry->d_inode;
+ struct nfs_server *nfss = NFS_SERVER(inode);
+ int status = 0;
+
+ *fsdata = lseg;
+ if (lseg && nfss->pnfs_curr_ld->write_begin)
+ status = _pnfs_write_begin(inode, page, pos, len, lseg,
+ (struct pnfs_fsdata **) fsdata);
+ return status;
+}
+
+static inline void pnfs_write_end_cleanup(struct file *filp, void *fsdata)
+{
+ struct nfs_server *nfss = NFS_SERVER(filp->f_dentry->d_inode);
+
+ if (fsdata && nfss->pnfs_curr_ld) {
+ if (nfss->pnfs_curr_ld->write_begin)
+ pnfs_free_fsdata(fsdata);
+ }
+}
+
static inline int pnfs_return_layout(struct inode *ino)
{
struct nfs_inode *nfsi = NFS_I(ino);
@@ -325,6 +363,19 @@ static inline void pnfs_pageio_init(struct nfs_pageio_descriptor *pgio,
pgio->pg_test = ld->pg_test;
}
+static inline struct pnfs_layout_segment *
+nfs4_pull_lseg_from_fsdata(struct file *filp, void *fsdata)
+{
+ if (fsdata) {
+ struct nfs_server *nfss = NFS_SERVER(filp->f_dentry->d_inode);
+
+ if (nfss->pnfs_curr_ld && nfss->pnfs_curr_ld->write_begin)
+ return ((struct pnfs_fsdata *) fsdata)->lseg;
+ return (struct pnfs_layout_segment *)fsdata;
+ }
+ return NULL;
+}
+
#else /* CONFIG_NFS_V4_1 */
static inline void pnfs_destroy_all_layouts(struct nfs_client *clp)
@@ -372,6 +423,19 @@ static inline int pnfs_return_layout(struct inode *ino)
return 0;
}
+static inline int pnfs_write_begin(struct file *filp, struct page *page,
+ loff_t pos, unsigned len,
+ struct pnfs_layout_segment *lseg,
+ void **fsdata)
+{
+ *fsdata = NULL;
+ return 0;
+}
+
+static inline void pnfs_write_end_cleanup(struct file *filp, void *fsdata)
+{
+}
+
static inline bool
pnfs_ld_layoutret_on_setattr(struct inode *inode)
{
@@ -443,6 +507,13 @@ static inline int pnfs_layoutcommit_inode(struct inode *inode, bool sync)
static inline void nfs4_deviceid_purge_client(struct nfs_client *ncl)
{
}
+
+static inline struct pnfs_layout_segment *
+nfs4_pull_lseg_from_fsdata(struct file *filp, void *fsdata)
+{
+ return NULL;
+}
+
#endif /* CONFIG_NFS_V4_1 */
#endif /* FS_NFS_PNFS_H */
--
1.7.4.1
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
