Re: Linux 2.6.7-rc2

[Peter Osterlund <petero2@telia.com>]

Linus Torvalds <torvalds@osdl.org> writes:

> On Mon, 31 May 2004, Peter Osterlund wrote:
> > 
> > If I put "#if 0" around the *wdata assignment in nfs_writepage_sync,
> > the stack usage goes down to 36, so it looks like gcc is building a
> > temporary structure on the stack and then copies the whole thing to
> > *wdata.
> 
> Yeah, that's silly. But understandable. A lot of problems go away by doing 
> a temporary private node..
> 
> > Does this construct save stack space for any version of gcc? Maybe the
> > code should be changed to do a memset() followed by explicit
> > initialization of the non-zero member variables instead.
> 
> In this case, I'd agree.
> 
> In some other cases, it's better to create a initialized static variable, 
> and just use that as an initial initializer. In this case that doesn't 
> much help, since none of the fields are constant.

Here is a patch that does this. I've verified that it compiles and
that it fixes the excessive stack problem, but I failed to come up
with a test case that exercises these code paths.


--- linux/fs/nfs/read.c.orig	2004-05-31 23:34:15.890307512 +0200
+++ linux/fs/nfs/read.c	2004-05-31 23:29:26.077365776 +0200
@@ -103,22 +103,16 @@
 	if (!rdata)
 		return -ENOMEM;
 
-	*rdata = (struct nfs_read_data) {
-		.flags		= (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0),
-		.cred		= NULL,
-		.inode		= inode,
-		.pages		= LIST_HEAD_INIT(rdata->pages),
-		.args		= {
-			.fh		= NFS_FH(inode),
-			.lockowner	= current->files,
-			.pages		= &page,
-			.pgbase		= 0UL,
-			.count		= rsize,
-		},
-		.res		= {
-			.fattr		= &rdata->fattr,
-		}
-	};
+	memset(rdata, 0, sizeof(*rdata));
+	rdata->flags = (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0);
+	rdata->inode = inode;
+	INIT_LIST_HEAD(&rdata->pages);
+	rdata->args.fh = NFS_FH(inode);
+	rdata->args.lockowner = current->files;
+	rdata->args.pages = &page;
+	rdata->args.pgbase = 0UL;
+	rdata->args.count = rsize;
+	rdata->res.fattr = &rdata->fattr;
 
 	dprintk("NFS: nfs_readpage_sync(%p)\n", page);
 
--- linux/fs/nfs/write.c.orig	2004-05-31 23:34:20.529602232 +0200
+++ linux/fs/nfs/write.c	2004-05-31 23:29:33.165288248 +0200
@@ -185,23 +185,17 @@
 	if (!wdata)
 		return -ENOMEM;
 
-	*wdata = (struct nfs_write_data) {
-		.flags		= how,
-		.cred		= NULL,
-		.inode		= inode,
-		.args		= {
-			.fh		= NFS_FH(inode),
-			.lockowner	= current->files,
-			.pages		= &page,
-			.stable		= NFS_FILE_SYNC,
-			.pgbase		= offset,
-			.count		= wsize,
-		},
-		.res		= {
-			.fattr		= &wdata->fattr,
-			.verf		= &wdata->verf,
-		},
-	};
+	memset(wdata, 0, sizeof(*wdata));
+	wdata->flags = how;
+	wdata->inode = inode;
+	wdata->args.fh = NFS_FH(inode);
+	wdata->args.lockowner = current->files;
+	wdata->args.pages = &page;
+	wdata->args.stable = NFS_FILE_SYNC;
+	wdata->args.pgbase = offset;
+	wdata->args.count = wsize;
+	wdata->res.fattr = &wdata->fattr;
+	wdata->res.verf = &wdata->verf;
 
 	dprintk("NFS:      nfs_writepage_sync(%s/%Ld %d@%Ld)\n",
 		inode->i_sb->s_id,

-- 
Peter Osterlund - petero2@telia.com
http://w1.894.telia.com/~u89404340