From:   Trond Myklebust <trondmy@gmail.com>
To:     linux-nfs@vger.kernel.org
Subject: [PATCH v2 23/28] NFS: Fix up NFS I/O subrequest creation
Date:   Fri, 29 Mar 2019 17:59:43 -0400
Message-Id: <20190329215948.107328-24-trond.myklebust@hammerspace.com>
In-Reply-To: <20190329215948.107328-23-trond.myklebust@hammerspace.com>
References: <20190329215948.107328-1-trond.myklebust@hammerspace.com>
 <20190329215948.107328-2-trond.myklebust@hammerspace.com>
 <20190329215948.107328-3-trond.myklebust@hammerspace.com>
 <20190329215948.107328-4-trond.myklebust@hammerspace.com>
 <20190329215948.107328-5-trond.myklebust@hammerspace.com>
 <20190329215948.107328-6-trond.myklebust@hammerspace.com>
 <20190329215948.107328-7-trond.myklebust@hammerspace.com>
 <20190329215948.107328-8-trond.myklebust@hammerspace.com>
 <20190329215948.107328-9-trond.myklebust@hammerspace.com>
 <20190329215948.107328-10-trond.myklebust@hammerspace.com>
 <20190329215948.107328-11-trond.myklebust@hammerspace.com>
 <20190329215948.107328-12-trond.myklebust@hammerspace.com>
 <20190329215948.107328-13-trond.myklebust@hammerspace.com>
 <20190329215948.107328-14-trond.myklebust@hammerspace.com>
 <20190329215948.107328-15-trond.myklebust@hammerspace.com>
 <20190329215948.107328-16-trond.myklebust@hammerspace.com>
 <20190329215948.107328-17-trond.myklebust@hammerspace.com>
 <20190329215948.107328-18-trond.myklebust@hammerspace.com>
 <20190329215948.107328-19-trond.myklebust@hammerspace.com>
 <20190329215948.107328-20-trond.myklebust@hammerspace.com>
 <20190329215948.107328-21-trond.myklebust@hammerspace.com>
 <20190329215948.107328-22-trond.myklebust@hammerspace.com>
 <20190329215948.107328-23-trond.myklebust@hammerspace.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Sender: linux-nfs-owner@vger.kernel.org
Precedence: bulk

We require all NFS I/O subrequests to duplicate the lock context as well
as the open context.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/pagelist.c | 93 ++++++++++++++++++++++++++++-------------------
 1 file changed, 55 insertions(+), 38 deletions(-)

diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c
index e9f39fa5964b..66a5c5d4a777 100644
--- a/fs/nfs/pagelist.c
+++ b/fs/nfs/pagelist.c
@@ -295,25 +295,13 @@ nfs_page_group_destroy(struct kref *kref)
 		nfs_release_request(head);
 }
 
-/**
- * nfs_create_request - Create an NFS read/write request.
- * @ctx: open context to use
- * @page: page to write
- * @last: last nfs request created for this page group or NULL if head
- * @offset: starting offset within the page for the write
- * @count: number of bytes to read/write
- *
- * The page must be locked by the caller. This makes sure we never
- * create two different requests for the same page.
- * User should ensure it is safe to sleep in this function.
- */
-struct nfs_page *
-nfs_create_request(struct nfs_open_context *ctx, struct page *page,
-		   struct nfs_page *last, unsigned int offset,
-		   unsigned int count)
+static struct nfs_page *
+__nfs_create_request(struct nfs_lock_context *l_ctx, struct page *page,
+		   struct nfs_page *last, unsigned int pgbase,
+		   unsigned int offset, unsigned int count)
 {
 	struct nfs_page		*req;
-	struct nfs_lock_context *l_ctx;
+	struct nfs_open_context *ctx = l_ctx->open_context;
 
 	if (test_bit(NFS_CONTEXT_BAD, &ctx->flags))
 		return ERR_PTR(-EBADF);
@@ -322,13 +310,8 @@ nfs_create_request(struct nfs_open_context *ctx, struct page *page,
 	if (req == NULL)
 		return ERR_PTR(-ENOMEM);
 
-	/* get lock context early so we can deal with alloc failures */
-	l_ctx = nfs_get_lock_context(ctx);
-	if (IS_ERR(l_ctx)) {
-		nfs_page_free(req);
-		return ERR_CAST(l_ctx);
-	}
 	req->wb_lock_context = l_ctx;
+	refcount_inc(&l_ctx->count);
 	atomic_inc(&l_ctx->io_count);
 
 	/* Initialize the request struct. Initially, we assume a
@@ -340,7 +323,7 @@ nfs_create_request(struct nfs_open_context *ctx, struct page *page,
 		get_page(page);
 	}
 	req->wb_offset  = offset;
-	req->wb_pgbase	= offset;
+	req->wb_pgbase	= pgbase;
 	req->wb_bytes   = count;
 	req->wb_context = get_nfs_open_context(ctx);
 	kref_init(&req->wb_kref);
@@ -348,6 +331,49 @@ nfs_create_request(struct nfs_open_context *ctx, struct page *page,
 	return req;
 }
 
+/**
+ * nfs_create_request - Create an NFS read/write request.
+ * @ctx: open context to use
+ * @page: page to write
+ * @last: last nfs request created for this page group or NULL if head
+ * @offset: starting offset within the page for the write
+ * @count: number of bytes to read/write
+ *
+ * The page must be locked by the caller. This makes sure we never
+ * create two different requests for the same page.
+ * User should ensure it is safe to sleep in this function.
+ */
+struct nfs_page *
+nfs_create_request(struct nfs_open_context *ctx, struct page *page,
+		   struct nfs_page *last, unsigned int offset,
+		   unsigned int count)
+{
+	struct nfs_lock_context *l_ctx = nfs_get_lock_context(ctx);
+	struct nfs_page *ret;
+
+	if (IS_ERR(l_ctx))
+		return ERR_CAST(l_ctx);
+	ret = __nfs_create_request(l_ctx, page, last, offset, offset, count);
+	nfs_put_lock_context(l_ctx);
+	return ret;
+}
+
+static struct nfs_page *
+nfs_create_subreq(struct nfs_page *req, struct nfs_page *last,
+		  unsigned int pgbase, unsigned int offset,
+		  unsigned int count)
+{
+	struct nfs_page *ret;
+
+	ret = __nfs_create_request(req->wb_lock_context, req->wb_page, last,
+			pgbase, offset, count);
+	if (!IS_ERR(ret)) {
+		nfs_lock_request(ret);
+		ret->wb_index = req->wb_index;
+	}
+	return ret;
+}
+
 /**
  * nfs_unlock_request - Unlock request and wake up sleepers.
  * @req: pointer to request
@@ -1049,14 +1075,10 @@ static int __nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
 		pgbase += subreq->wb_bytes;
 
 		if (bytes_left) {
-			subreq = nfs_create_request(req->wb_context,
-					req->wb_page,
-					subreq, pgbase, bytes_left);
+			subreq = nfs_create_subreq(req, subreq, pgbase,
+					offset, bytes_left);
 			if (IS_ERR(subreq))
 				goto err_ptr;
-			nfs_lock_request(subreq);
-			subreq->wb_offset  = offset;
-			subreq->wb_index = req->wb_index;
 		}
 	} while (bytes_left > 0);
 
@@ -1158,19 +1180,14 @@ int nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
 			     lastreq = lastreq->wb_this_page)
 				;
 
-			dupreq = nfs_create_request(req->wb_context,
-					req->wb_page, lastreq, pgbase, bytes);
+			dupreq = nfs_create_subreq(req, lastreq,
+					pgbase, offset, bytes);
 
+			nfs_page_group_unlock(req);
 			if (IS_ERR(dupreq)) {
-				nfs_page_group_unlock(req);
 				desc->pg_error = PTR_ERR(dupreq);
 				goto out_failed;
 			}
-
-			nfs_lock_request(dupreq);
-			nfs_page_group_unlock(req);
-			dupreq->wb_offset = offset;
-			dupreq->wb_index = req->wb_index;
 		} else
 			dupreq = req;
 
-- 
2.20.1