1 files changed, 264 insertions, 155 deletions

diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 874972d9427c..727168059684 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
 
@@ -55,10 +56,10 @@ struct nfs_write_data *nfs_commitdata_alloc(void)
         if (p) {
                 memset(p, 0, sizeof(*p));
                 INIT_LIST_HEAD(&p->pages);
-                p->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
         }
         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)
 {
@@ -75,7 +77,6 @@ struct nfs_write_data *nfs_writedata_alloc(unsigned int pagecount)
                 memset(p, 0, sizeof(*p));
                 INIT_LIST_HEAD(&p->pages);
                 p->npages = pagecount;
-                p->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
                 if (pagecount <= ARRAY_SIZE(p->page_array))
                         p->pagevec = p->page_array;
                 else {
@@ -98,6 +99,7 @@ void nfs_writedata_free(struct nfs_write_data *p)
 
 static void nfs_writedata_release(struct nfs_write_data *wdata)
 {
+        put_lseg(wdata->lseg);
         put_nfs_open_context(wdata->args.context);
         nfs_writedata_free(wdata);
 }
@@ -179,8 +181,8 @@ 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;
 }
 
 /*
@@ -292,9 +294,7 @@ static int nfs_do_writepage(struct page *page, struct writeback_control *wbc, st
         nfs_add_stats(inode, NFSIOS_WRITEPAGES, 1);
 
         nfs_pageio_cond_complete(pgio, page->index);
-        ret = nfs_page_async_flush(pgio, page,
-                        wbc->sync_mode == WB_SYNC_NONE ||
-                        wbc->nonblocking != 0);
+        ret = nfs_page_async_flush(pgio, page, wbc->sync_mode == WB_SYNC_NONE);
         if (ret == -EAGAIN) {
                 redirty_page_for_writepage(wbc, page);
                 ret = 0;
@@ -389,11 +389,9 @@ static int nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
         spin_lock(&inode->i_lock);
         error = radix_tree_insert(&nfsi->nfs_page_tree, req->wb_index, req);
         BUG_ON(error);
-        if (!nfsi->npages) {
-                igrab(inode);
-                if (nfs_have_delegation(inode, FMODE_WRITE))
-                        nfsi->change_attr++;
-        }
+        if (!nfsi->npages && nfs_have_delegation(inode, FMODE_WRITE))
+                nfsi->change_attr++;
+        set_bit(PG_MAPPED, &req->wb_flags);
         SetPagePrivate(req->wb_page);
         set_page_private(req->wb_page, (unsigned long)req);
         nfsi->npages++;
@@ -419,14 +417,10 @@ static void nfs_inode_remove_request(struct nfs_page *req)
         spin_lock(&inode->i_lock);
         set_page_private(req->wb_page, 0);
         ClearPagePrivate(req->wb_page);
+        clear_bit(PG_MAPPED, &req->wb_flags);
         radix_tree_delete(&nfsi->nfs_page_tree, req->wb_index);
         nfsi->npages--;
-        if (!nfsi->npages) {
-                spin_unlock(&inode->i_lock);
-                iput(inode);
-        } else
-                spin_unlock(&inode->i_lock);
-        nfs_clear_request(req);
+        spin_unlock(&inode->i_lock);
         nfs_release_request(req);
 }
 
@@ -442,7 +436,7 @@ 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);
@@ -454,6 +448,7 @@ 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);
@@ -475,14 +470,18 @@ 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
-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)) {
@@ -493,7 +492,7 @@ 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)
 {
 }
 
@@ -510,7 +509,8 @@ 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;
 }
@@ -542,11 +542,15 @@ nfs_scan_commit(struct inode *inode, struct list_head *dst, pgoff_t idx_start, u
         if (!nfs_need_commit(nfsi))
                 return 0;
 
+        spin_lock(&inode->i_lock);
         ret = nfs_scan_list(nfsi, dst, idx_start, npages, NFS_PAGE_TAG_COMMIT);
         if (ret > 0)
                 nfsi->ncommit -= ret;
+        spin_unlock(&inode->i_lock);
+
         if (nfs_need_commit(NFS_I(inode)))
                 __mark_inode_dirty(inode, I_DIRTY_DATASYNC);
+
         return ret;
 }
 #else
@@ -613,9 +617,11 @@ 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) {
@@ -674,7 +680,6 @@ static int nfs_writepage_setup(struct nfs_open_context *ctx, struct page *page,
         req = nfs_setup_write_request(ctx, page, offset, count);
         if (IS_ERR(req))
                 return PTR_ERR(req);
-        nfs_mark_request_dirty(req);
         /* Update file length */
         nfs_grow_file(page, offset, count);
         nfs_mark_uptodate(page, req->wb_pgbase, req->wb_bytes);
@@ -763,11 +768,12 @@ 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);
@@ -784,25 +790,21 @@ static int flush_task_priority(int how)
         return RPC_PRIORITY_NORMAL;
 }
 
-/*
- * Set up the argument/result storage required for the RPC call.
- */
-static int nfs_write_rpcsetup(struct nfs_page *req,
-                struct nfs_write_data *data,
-                const struct rpc_call_ops *call_ops,
-                unsigned int count, unsigned int offset,
-                int how)
+int nfs_initiate_write(struct nfs_write_data *data,
+                       struct rpc_clnt *clnt,
+                       const struct rpc_call_ops *call_ops,
+                       int how)
 {
-        struct inode *inode = req->wb_context->path.dentry->d_inode;
+        struct inode *inode = data->inode;
         int priority = flush_task_priority(how);
         struct rpc_task *task;
         struct rpc_message msg = {
                 .rpc_argp = &data->args,
                 .rpc_resp = &data->res,
-                .rpc_cred = req->wb_context->cred,
+                .rpc_cred = data->cred,
         };
         struct rpc_task_setup task_setup_data = {
-                .rpc_client = NFS_CLIENT(inode),
+                .rpc_client = clnt,
                 .task = &data->task,
                 .rpc_message = &msg,
                 .callback_ops = call_ops,
@@ -813,22 +815,64 @@ static int nfs_write_rpcsetup(struct nfs_page *req,
         };
         int ret = 0;
 
+        /* Set up the initial task struct.  */
+        NFS_PROTO(inode)->write_setup(data, &msg);
+
+        dprintk("NFS: %5u initiated write call "
+                "(req %s/%lld, %u bytes @ offset %llu)\n",
+                data->task.tk_pid,
+                inode->i_sb->s_id,
+                (long long)NFS_FILEID(inode),
+                data->args.count,
+                (unsigned long long)data->args.offset);
+
+        task = rpc_run_task(&task_setup_data);
+        if (IS_ERR(task)) {
+                ret = PTR_ERR(task);
+                goto out;
+        }
+        if (how & FLUSH_SYNC) {
+                ret = rpc_wait_for_completion_task(task);
+                if (ret == 0)
+                        ret = task->tk_status;
+        }
+        rpc_put_task(task);
+out:
+        return ret;
+}
+EXPORT_SYMBOL_GPL(nfs_initiate_write);
+
+/*
+ * Set up the argument/result storage required for the RPC call.
+ */
+static int nfs_write_rpcsetup(struct nfs_page *req,
+                struct nfs_write_data *data,
+                const struct rpc_call_ops *call_ops,
+                unsigned int count, unsigned int offset,
+                struct pnfs_layout_segment *lseg,
+                int how)
+{
+        struct inode *inode = req->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. */
 
         data->req = req;
         data->inode = inode = req->wb_context->path.dentry->d_inode;
-        data->cred = msg.rpc_cred;
+        data->cred = req->wb_context->cred;
+        data->lseg = get_lseg(lseg);
 
         data->args.fh     = NFS_FH(inode);
         data->args.offset = req_offset(req) + offset;
+        /* pnfs_set_layoutcommit needs this */
+        data->mds_offset = data->args.offset;
         data->args.pgbase = req->wb_pgbase + offset;
         data->args.pages  = data->pagevec;
         data->args.count  = count;
         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;
@@ -839,30 +883,11 @@ static int nfs_write_rpcsetup(struct nfs_page *req,
         data->res.verf    = &data->verf;
         nfs_fattr_init(&data->fattr);
 
-        /* Set up the initial task struct.  */
-        NFS_PROTO(inode)->write_setup(data, &msg);
-
-        dprintk("NFS: %5u initiated write call "
-                "(req %s/%lld, %u bytes @ offset %llu)\n",
-                data->task.tk_pid,
-                inode->i_sb->s_id,
-                (long long)NFS_FILEID(inode),
-                count,
-                (unsigned long long)data->args.offset);
+        if (data->lseg &&
+            (pnfs_try_to_write_data(data, call_ops, how) == PNFS_ATTEMPTED))
+                return 0;
 
-        task = rpc_run_task(&task_setup_data);
-        if (IS_ERR(task)) {
-                ret = PTR_ERR(task);
-                goto out;
-        }
-        if (how & FLUSH_SYNC) {
-                ret = rpc_wait_for_completion_task(task);
-                if (ret == 0)
-                        ret = task->tk_status;
-        }
-        rpc_put_task(task);
-out:
-        return ret;
+        return nfs_initiate_write(data, NFS_CLIENT(inode), call_ops, how);
 }
 
 /* If a nfs_flush_* function fails, it should remove reqs from @head and
@@ -882,20 +907,27 @@ static void nfs_redirty_request(struct nfs_page *req)
  * Generate multiple small requests to write out a single
  * contiguous dirty area on one page.
  */
-static int nfs_flush_multi(struct inode *inode, struct list_head *head, unsigned int npages, size_t count, int how)
+static int nfs_flush_multi(struct nfs_pageio_descriptor *desc)
 {
-        struct nfs_page *req = nfs_list_entry(head->next);
+        struct nfs_page *req = nfs_list_entry(desc->pg_list.next);
         struct page *page = req->wb_page;
         struct nfs_write_data *data;
-        size_t wsize = NFS_SERVER(inode)->wsize, nbytes;
+        size_t wsize = NFS_SERVER(desc->pg_inode)->wsize, nbytes;
         unsigned int offset;
         int requests = 0;
         int ret = 0;
+        struct pnfs_layout_segment *lseg;
         LIST_HEAD(list);
 
         nfs_list_remove_request(req);
 
-        nbytes = count;
+        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);
 
@@ -908,9 +940,13 @@ static int nfs_flush_multi(struct inode *inode, struct list_head *head, unsigned
         } while (nbytes != 0);
         atomic_set(&req->wb_complete, requests);
 
+        BUG_ON(desc->pg_lseg);
+        lseg = pnfs_update_layout(desc->pg_inode, req->wb_context,
+                                  req_offset(req), desc->pg_count,
+                                  IOMODE_RW, GFP_NOFS);
         ClearPageError(page);
         offset = 0;
-        nbytes = count;
+        nbytes = desc->pg_count;
         do {
                 int ret2;
 
@@ -922,20 +958,22 @@ static int nfs_flush_multi(struct inode *inode, struct list_head *head, unsigned
                 if (nbytes < wsize)
                         wsize = nbytes;
                 ret2 = nfs_write_rpcsetup(req, data, &nfs_write_partial_ops,
-                                   wsize, offset, how);
+                                          wsize, offset, lseg, desc->pg_ioflags);
                 if (ret == 0)
                         ret = ret2;
                 offset += wsize;
                 nbytes -= wsize;
         } while (nbytes != 0);
 
+        put_lseg(lseg);
+        desc->pg_lseg = NULL;
         return ret;
 
 out_bad:
         while (!list_empty(&list)) {
                 data = list_entry(list.next, struct nfs_write_data, pages);
                 list_del(&data->pages);
-                nfs_writedata_release(data);
+                nfs_writedata_free(data);
         }
         nfs_redirty_request(req);
         return -ENOMEM;
@@ -949,16 +987,26 @@ out_bad:
  * This is the case if nfs_updatepage detects a conflicting request
  * that has been written but not committed.
  */
-static int nfs_flush_one(struct inode *inode, struct list_head *head, unsigned int npages, size_t count, int how)
+static int nfs_flush_one(struct nfs_pageio_descriptor *desc)
 {
         struct nfs_page         *req;
         struct page             **pages;
         struct nfs_write_data   *data;
+        struct list_head *head = &desc->pg_list;
+        struct pnfs_layout_segment *lseg = desc->pg_lseg;
+        int ret;
 
-        data = nfs_writedata_alloc(npages);
-        if (!data)
-                goto out_bad;
-
+        data = nfs_writedata_alloc(nfs_page_array_len(desc->pg_base,
+                                                      desc->pg_count));
+        if (!data) {
+                while (!list_empty(head)) {
+                        req = nfs_list_entry(head->next);
+                        nfs_list_remove_request(req);
+                        nfs_redirty_request(req);
+                }
+                ret = -ENOMEM;
+                goto out;
+        }
         pages = data->pagevec;
         while (!list_empty(head)) {
                 req = nfs_list_entry(head->next);
@@ -968,16 +1016,21 @@ static int nfs_flush_one(struct inode *inode, struct list_head *head, unsigned i
                 *pages++ = req->wb_page;
         }
         req = nfs_list_entry(data->pages.next);
+        if ((!lseg) && list_is_singular(&data->pages))
+                lseg = pnfs_update_layout(desc->pg_inode, req->wb_context,
+                                          req_offset(req), desc->pg_count,
+                                          IOMODE_RW, GFP_NOFS);
+
+        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 */
-        return nfs_write_rpcsetup(req, data, &nfs_write_full_ops, count, 0, how);
- out_bad:
-        while (!list_empty(head)) {
-                req = nfs_list_entry(head->next);
-                nfs_list_remove_request(req);
-                nfs_redirty_request(req);
-        }
-        return -ENOMEM;
+        ret = nfs_write_rpcsetup(req, data, &nfs_write_full_ops, desc->pg_count, 0, lseg, desc->pg_ioflags);
+out:
+        put_lseg(lseg); /* Cleans any gotten in ->pg_test */
+        desc->pg_lseg = NULL;
+        return ret;
 }
 
 static void nfs_pageio_init_write(struct nfs_pageio_descriptor *pgio,
@@ -1042,7 +1095,7 @@ 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);
 }
 
@@ -1109,7 +1162,7 @@ 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;
                 }
@@ -1135,7 +1188,7 @@ static const struct rpc_call_ops nfs_write_full_ops = {
 /*
  * This function is called when the WRITE call is complete.
  */
-int nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
+void nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
 {
         struct nfs_writeargs    *argp = &data->args;
         struct nfs_writeres     *resp = &data->res;
@@ -1154,7 +1207,7 @@ int nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
          */
         status = NFS_PROTO(data->inode)->write_done(task, data);
         if (status != 0)
-                return status;
+                return;
         nfs_add_stats(data->inode, NFSIOS_SERVERWRITTENBYTES, resp->count);
 
 #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
@@ -1169,6 +1222,7 @@ int nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
                  */
                 static unsigned long    complain;
 
+                /* Note this will print the MDS for a DS write */
                 if (time_before(complain, jiffies)) {
                         dprintk("NFS:       faulty NFS server %s:"
                                 " (committed = %d) != (stable = %d)\n",
@@ -1189,6 +1243,7 @@ int nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
                         /* Was this an NFSv2 write or an NFSv3 stable write? */
                         if (resp->verf->committed != NFS_UNSTABLE) {
                                 /* Resend from where the server left off */
+                                data->mds_offset += resp->count;
                                 argp->offset += resp->count;
                                 argp->pgbase += resp->count;
                                 argp->count -= resp->count;
@@ -1199,7 +1254,7 @@ int nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
                                 argp->stable = NFS_FILE_SYNC;
                         }
                         nfs_restart_rpc(task, server->nfs_client);
-                        return -EAGAIN;
+                        return;
                 }
                 if (time_before(complain, jiffies)) {
                         printk(KERN_WARNING
@@ -1210,64 +1265,89 @@ int nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
                 /* Can't do anything about it except throw an error. */
                 task->tk_status = -EIO;
         }
-        return 0;
+        return;
 }
 
 
 #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)
+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;
 
+        put_lseg(wdata->lseg);
         put_nfs_open_context(wdata->args.context);
         nfs_commit_free(wdata);
 }
+EXPORT_SYMBOL_GPL(nfs_commitdata_release);
 
-/*
- * 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)
+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;
+}
+EXPORT_SYMBOL_GPL(nfs_initiate_commit);
+
+/*
+ * Set up the argument/result storage required for the RPC call.
+ */
+void nfs_init_commit(struct nfs_write_data *data,
+                            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;
 
         /* Set up the RPC argument and reply structs
          * NB: take care not to mess about with data->commit et al. */
@@ -1275,7 +1355,9 @@ 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->lseg        = lseg; /* reference transferred */
+        data->mds_ops     = &nfs_commit_ops;
 
         data->args.fh     = NFS_FH(data->inode);
         /* Note: we always request a commit of the entire inode */
@@ -1286,18 +1368,25 @@ static int nfs_commit_rpcsetup(struct list_head *head,
         data->res.fattr   = &data->fattr;
         data->res.verf    = &data->verf;
         nfs_fattr_init(&data->fattr);
+}
+EXPORT_SYMBOL_GPL(nfs_init_commit);
 
-        /* Set up the initial task struct.  */
-        NFS_PROTO(inode)->commit_setup(data, &msg);
-
-        dprintk("NFS: %5u initiated commit call\n", data->task.tk_pid);
+void nfs_retry_commit(struct list_head *page_list,
+                      struct pnfs_layout_segment *lseg)
+{
+        struct nfs_page *req;
 
-        task = rpc_run_task(&task_setup_data);
-        if (IS_ERR(task))
-                return PTR_ERR(task);
-        rpc_put_task(task);
-        return 0;
+        while (!list_empty(page_list)) {
+                req = nfs_list_entry(page_list->next);
+                nfs_list_remove_request(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);
+                nfs_clear_page_tag_locked(req);
+        }
 }
+EXPORT_SYMBOL_GPL(nfs_retry_commit);
 
 /*
  * Commit dirty pages
@@ -1306,7 +1395,6 @@ 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();
 
@@ -1314,17 +1402,10 @@ 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, NULL);
+        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, NULL);
         nfs_commit_clear_lock(NFS_I(inode));
         return -ENOMEM;
 }
@@ -1340,14 +1421,12 @@ static void nfs_commit_done(struct rpc_task *task, void *calldata)
                                 task->tk_pid, task->tk_status);
 
         /* Call the NFS version-specific code */
-        if (NFS_PROTO(data->inode)->commit_done(task, data) != 0)
-                return;
+        NFS_PROTO(data->inode)->commit_done(task, data);
 }
 
-static void nfs_commit_release(void *calldata)
+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)) {
@@ -1381,6 +1460,14 @@ static void nfs_commit_release(void *calldata)
         next:
                 nfs_clear_page_tag_locked(req);
         }
+}
+EXPORT_SYMBOL_GPL(nfs_commit_release_pages);
+
+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);
 }
@@ -1397,23 +1484,28 @@ 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);
-        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)
-                        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;
@@ -1433,15 +1525,17 @@ static int nfs_commit_unstable_pages(struct inode *inode, struct writeback_contr
         int flags = FLUSH_SYNC;
         int ret = 0;
 
-        /* Don't commit yet if this is a non-blocking flush and there are
-         * lots of outstanding writes for this mapping.
-         */
-        if (wbc->sync_mode == WB_SYNC_NONE &&
-            nfsi->ncommit <= (nfsi->npages >> 1))
-                goto out_mark_dirty;
+        if (wbc->sync_mode == WB_SYNC_NONE) {
+                /* Don't commit yet if this is a non-blocking flush and there
+                 * are a lot of outstanding writes for this mapping.
+                 */
+                if (nfsi->ncommit <= (nfsi->npages >> 1))
+                        goto out_mark_dirty;
 
-        if (wbc->nonblocking || wbc->for_background)
+                /* don't wait for the COMMIT response */
                 flags = 0;
+        }
+
         ret = nfs_commit_inode(inode, flags);
         if (ret >= 0) {
                 if (wbc->sync_mode == WB_SYNC_NONE) {
@@ -1465,7 +1559,22 @@ 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;
+                bool sync = true;
+
+                if (wbc->sync_mode == WB_SYNC_NONE || wbc->nonblocking ||
+                    wbc->for_background)
+                        sync = false;
+
+                status = pnfs_layoutcommit_inode(inode, sync);
+                if (status < 0)
+                        return status;
+        }
+        return ret;
 }
 
 /*