NFSv4.1: coelesce across layout stripes
Add a pg_test layout driver hook which is used to avoid coelescing I/O across
layout stripes.
Signed-off-by: Andy Adamson <andros@netapp.com>
Signed-off-by: Andy Adamson <andros@citi.umich.edu>
Signed-off-by: Dean Hildebrand <dhildeb@us.ibm.com>
Signed-off-by: Fred Isaman <iisaman@citi.umich.edu>
Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Signed-off-by: Oleg Drokin <green@linuxhacker.ru>
Signed-off-by: Tao Guo <guotao@nrchpc.ac.cn>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c
index 23f930c..0efe8cb 100644
--- a/fs/nfs/nfs4filelayout.c
+++ b/fs/nfs/nfs4filelayout.c
@@ -252,6 +252,31 @@
_filelayout_free_lseg(fl);
}
+/*
+ * filelayout_pg_test(). Called by nfs_can_coalesce_requests()
+ *
+ * return 1 : coalesce page
+ * return 0 : don't coalesce page
+ */
+int
+filelayout_pg_test(struct nfs_pageio_descriptor *pgio, struct nfs_page *prev,
+ struct nfs_page *req)
+{
+ u64 p_stripe, r_stripe;
+ u32 stripe_unit;
+
+ if (!pgio->pg_lseg)
+ return 1;
+ p_stripe = (u64)prev->wb_index << PAGE_CACHE_SHIFT;
+ r_stripe = (u64)req->wb_index << PAGE_CACHE_SHIFT;
+ stripe_unit = FILELAYOUT_LSEG(pgio->pg_lseg)->stripe_unit;
+
+ do_div(p_stripe, stripe_unit);
+ do_div(r_stripe, stripe_unit);
+
+ return (p_stripe == r_stripe);
+}
+
static struct pnfs_layoutdriver_type filelayout_type = {
.id = LAYOUT_NFSV4_1_FILES,
.name = "LAYOUT_NFSV4_1_FILES",
@@ -260,6 +285,7 @@
.clear_layoutdriver = filelayout_clear_layoutdriver,
.alloc_lseg = filelayout_alloc_lseg,
.free_lseg = filelayout_free_lseg,
+ .pg_test = filelayout_pg_test,
};
static int __init nfs4filelayout_init(void)
diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c
index e1164e3..9b9a65c 100644
--- a/fs/nfs/pagelist.c
+++ b/fs/nfs/pagelist.c
@@ -226,6 +226,7 @@
desc->pg_doio = doio;
desc->pg_ioflags = io_flags;
desc->pg_error = 0;
+ desc->pg_lseg = NULL;
}
/**
@@ -240,7 +241,8 @@
* Return 'true' if this is the case, else return 'false'.
*/
static int nfs_can_coalesce_requests(struct nfs_page *prev,
- struct nfs_page *req)
+ struct nfs_page *req,
+ struct nfs_pageio_descriptor *pgio)
{
if (req->wb_context->cred != prev->wb_context->cred)
return 0;
@@ -254,6 +256,12 @@
return 0;
if (prev->wb_pgbase + prev->wb_bytes != PAGE_CACHE_SIZE)
return 0;
+ /*
+ * Non-whole file layouts need to check that req is inside of
+ * pgio->pg_lseg.
+ */
+ if (pgio->pg_test && !pgio->pg_test(pgio, prev, req))
+ return 0;
return 1;
}
@@ -286,7 +294,7 @@
if (newlen > desc->pg_bsize)
return 0;
prev = nfs_list_entry(desc->pg_list.prev);
- if (!nfs_can_coalesce_requests(prev, req))
+ if (!nfs_can_coalesce_requests(prev, req, desc))
return 0;
} else
desc->pg_base = req->wb_pgbase;
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 3afa82e..330cee1 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -858,6 +858,22 @@
goto out;
}
+static void
+pnfs_set_pg_test(struct inode *inode, struct nfs_pageio_descriptor *pgio)
+{
+ struct pnfs_layoutdriver_type *ld;
+
+ ld = NFS_SERVER(inode)->pnfs_curr_ld;
+ pgio->pg_test = (ld ? ld->pg_test : NULL);
+}
+
+void
+pnfs_pageio_init_read(struct nfs_pageio_descriptor *pgio,
+ struct inode *inode)
+{
+ pnfs_set_pg_test(inode, pgio);
+}
+
/*
* Device ID cache. Currently supports one layout type per struct nfs_client.
* Add layout type to the lookup key to expand to support multiple types.
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index 9a994bc..db52d96 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -30,6 +30,8 @@
#ifndef FS_NFS_PNFS_H
#define FS_NFS_PNFS_H
+#include <linux/nfs_page.h>
+
enum {
NFS_LSEG_VALID = 0, /* cleared when lseg is recalled/returned */
NFS_LSEG_ROC, /* roc bit received from server */
@@ -65,6 +67,9 @@
int (*clear_layoutdriver) (struct nfs_server *);
struct pnfs_layout_segment * (*alloc_lseg) (struct pnfs_layout_hdr *layoutid, struct nfs4_layoutget_res *lgr);
void (*free_lseg) (struct pnfs_layout_segment *lseg);
+
+ /* test for nfs page cache coalescing */
+ int (*pg_test)(struct nfs_pageio_descriptor *, struct nfs_page *, struct nfs_page *);
};
struct pnfs_layout_hdr {
@@ -151,6 +156,7 @@
enum pnfs_iomode access_type);
void set_pnfs_layoutdriver(struct nfs_server *, u32 id);
void unset_pnfs_layoutdriver(struct nfs_server *);
+void pnfs_pageio_init_read(struct nfs_pageio_descriptor *, struct inode *);
int pnfs_layout_process(struct nfs4_layoutget *lgp);
void pnfs_free_lseg_list(struct list_head *tmp_list);
void pnfs_destroy_layout(struct nfs_inode *);
@@ -250,6 +256,12 @@
{
}
+static inline void
+pnfs_pageio_init_read(struct nfs_pageio_descriptor *pgio, struct inode *ino)
+{
+ pgio->pg_test = NULL;
+}
+
#endif /* CONFIG_NFS_V4_1 */
#endif /* FS_NFS_PNFS_H */
diff --git a/fs/nfs/read.c b/fs/nfs/read.c
index aedcaa7..2a27659 100644
--- a/fs/nfs/read.c
+++ b/fs/nfs/read.c
@@ -626,6 +626,7 @@
goto read_complete; /* all pages were read */
pnfs_update_layout(inode, desc.ctx, IOMODE_READ);
+ pnfs_pageio_init_read(&pgio, inode);
if (rsize < PAGE_CACHE_SIZE)
nfs_pageio_init(&pgio, inode, nfs_pagein_multi, rsize, 0);
else
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index ae528b9..40143c4 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -28,6 +28,7 @@
#include "iostat.h"
#include "nfs4_fs.h"
#include "fscache.h"
+#include "pnfs.h"
#define NFSDBG_FACILITY NFSDBG_PAGECACHE
@@ -982,6 +983,8 @@
{
size_t wsize = NFS_SERVER(inode)->wsize;
+ pgio->pg_test = NULL;
+
if (wsize < PAGE_CACHE_SIZE)
nfs_pageio_init(pgio, inode, nfs_flush_multi, wsize, ioflags);
else