From: Wendy Cheng <wcheng@redhat.com>
Subject: [RHEL5.1 PATCH] 3/4 BZ 243136 GFS2 fh imode passing
Date: Thu, 28 Jun 2007 01:28:54 -0400
Bugzilla: 243136
Message-Id: <46834716.405@redhat.com>
Changelog: [gfs2] Remove i_mode passing from NFS File Handle


BZ 243136 [patch 3/4] Remove i_mode passing from NFS File Handle

GFS2 passes i_mode within NFS File Handle. Other than the wrong assumption
that this extra 16 bit value can always be added into file handle, the NFSD
get dentry doesn't really need the i_mode to work correctly. It goes thru 
the same lookup code path as direct file access route (where the mode is 
obtained from name lookup) during lookup time. Upon file access procedure 
call, if on-disk inode is not in-memory, it will read-in the disk inode as
part of the glock operation where imode can be obtained. This makes i_mode 
inside File Handle an obsolete feature. The patch removes it accordingly. 
Without the removal, current NFS over GFS2 will panic with deadlock assert. 

-- Wendy


--- gfs2-nfs2/fs/gfs2/inode.h	2007-06-27 00:17:41.000000000 -0400
+++ gfs2-nfs/fs/gfs2/inode.h	2007-06-27 14:56:46.000000000 -0400
@@ -47,6 +47,7 @@ static inline void gfs2_inum_out(const s
 
 
 void gfs2_inode_attr_in(struct gfs2_inode *ip);
+void gfs2_set_iop(struct inode *inode);
 struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned type,
 				u64 no_addr, u64 no_formal_ino);
 struct inode *gfs2_ilookup(struct super_block *sb, u64 no_addr);
--- gfs2-nfs2/fs/gfs2/inode.c	2007-06-27 01:17:18.000000000 -0400
+++ gfs2-nfs/fs/gfs2/inode.c	2007-06-27 14:59:48.000000000 -0400
@@ -78,6 +78,36 @@ static struct inode *gfs2_iget(struct su
 }
 
 /**
+ * GFS2 lookup code fills in vfs inode contents based on info obtained
+ * from directory entry inside gfs2_inode_lookup(). This has caused issues
+ * with NFS code path since its get_dentry routine doesn't have the relevant
+ * directory entry when gfs2_inode_lookup() is invoked. Part of the code
+ * segment inside gfs2_inode_lookup code needs to get moved around.
+ *
+ * Clean up I_LOCK and I_NEW as well.
+ **/
+
+void gfs2_set_iop(struct inode *inode)
+{
+	umode_t mode = inode->i_mode;
+
+	if (S_ISREG(mode)) {
+		inode->i_op = &gfs2_file_iops;
+		inode->i_fop = &gfs2_file_fops;
+		inode->i_mapping->a_ops = &gfs2_file_aops;
+	} else if (S_ISDIR(mode)) {
+		inode->i_op = &gfs2_dir_iops;
+		inode->i_fop = &gfs2_dir_fops;
+	} else if (S_ISLNK(mode)) {
+		inode->i_op = &gfs2_symlink_iops;
+	} else {
+		inode->i_op = &gfs2_dev_iops;
+	}
+
+	unlock_new_inode(inode);
+}
+
+/**
  * gfs2_inode_lookup - Lookup an inode
  * @sb: The super block
  * @no_addr: The inode number
@@ -101,7 +131,6 @@ struct inode *gfs2_inode_lookup(struct s
 
 	if (inode->i_state & I_NEW) {
 		struct gfs2_sbd *sdp = GFS2_SB(inode);
-		umode_t mode;
 		inode->i_private = ip;
 		ip->i_no_formal_ino = no_formal_ino;
 
@@ -122,6 +151,11 @@ struct inode *gfs2_inode_lookup(struct s
 
 		gfs2_glock_put(io_gl);
 
+		if ((type == DT_UNKNOWN) && (no_formal_ino == 0))
+			goto gfs2_nfsbypass;
+
+		inode->i_mode = DT2IF(type);
+
 		/*
 		 * We must read the inode in order to work out its type in
 		 * this case. Note that this doesn't happen often as we normally
@@ -129,33 +163,19 @@ struct inode *gfs2_inode_lookup(struct s
 		 * unlinked inode recovery (where it is safe to do this glock,
 		 * which is not true in the general case).
 		 */
-		inode->i_mode = mode = DT2IF(type);
 		if (type == DT_UNKNOWN) {
 			struct gfs2_holder gh;
 			error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
 			if (unlikely(error))
 				goto fail_glock;
 			/* Inode is now uptodate */
-			mode = inode->i_mode;
 			gfs2_glock_dq_uninit(&gh);
 		}
 
-		if (S_ISREG(mode)) {
-			inode->i_op = &gfs2_file_iops;
-			inode->i_fop = &gfs2_file_fops;
-			inode->i_mapping->a_ops = &gfs2_file_aops;
-		} else if (S_ISDIR(mode)) {
-			inode->i_op = &gfs2_dir_iops;
-			inode->i_fop = &gfs2_dir_fops;
-		} else if (S_ISLNK(mode)) {
-			inode->i_op = &gfs2_symlink_iops;
-		} else {
-			inode->i_op = &gfs2_dev_iops;
-		}
-
-		unlock_new_inode(inode);
+		gfs2_set_iop(inode);
 	}
 
+gfs2_nfsbypass:
 	return inode;
 fail_glock:
 	gfs2_glock_dq(&ip->i_iopen_gh);
--- gfs2-nfs2/fs/gfs2/ops_export.c	2007-06-27 00:24:38.000000000 -0400
+++ gfs2-nfs/fs/gfs2/ops_export.c	2007-06-27 14:57:05.000000000 -0400
@@ -27,12 +27,7 @@
 #include "util.h"
 
 #define GFS2_SMALL_FH_SIZE 4
-#define GFS2_LARGE_FH_SIZE 10
-
-struct gfs2_fh_obj {
-	struct gfs2_inum_host this;
-	u32 imode;
-};
+#define GFS2_LARGE_FH_SIZE 8
 
 static struct dentry *gfs2_decode_fh(struct super_block *sb,
 				     __u32 *p,
@@ -43,11 +38,8 @@ static struct dentry *gfs2_decode_fh(str
 				     void *context)
 {
 	__be32 *fh = (__force __be32 *)p;
-	struct gfs2_fh_obj fh_obj;
-	struct gfs2_inum_host *this, parent;
+	struct gfs2_inum_host inum, parent;
 
-	this 		= &fh_obj.this;
-	fh_obj.imode 	= DT_UNKNOWN;
 	memset(&parent, 0, sizeof(struct gfs2_inum));
 
 	switch (fh_len) {
@@ -56,18 +48,17 @@ static struct dentry *gfs2_decode_fh(str
 		parent.no_formal_ino |= be32_to_cpu(fh[5]);
 		parent.no_addr = ((u64)be32_to_cpu(fh[6])) << 32;
 		parent.no_addr |= be32_to_cpu(fh[7]);
-		fh_obj.imode = be32_to_cpu(fh[8]);
 	case GFS2_SMALL_FH_SIZE:
-		this->no_formal_ino = ((u64)be32_to_cpu(fh[0])) << 32;
-		this->no_formal_ino |= be32_to_cpu(fh[1]);
-		this->no_addr = ((u64)be32_to_cpu(fh[2])) << 32;
-		this->no_addr |= be32_to_cpu(fh[3]);
+		inum.no_formal_ino = ((u64)be32_to_cpu(fh[0])) << 32;
+		inum.no_formal_ino |= be32_to_cpu(fh[1]);
+		inum.no_addr = ((u64)be32_to_cpu(fh[2])) << 32;
+		inum.no_addr |= be32_to_cpu(fh[3]);
 		break;
 	default:
 		return NULL;
 	}
 
-	return gfs2_export_ops.find_exported_dentry(sb, &fh_obj, &parent,
+	return gfs2_export_ops.find_exported_dentry(sb, &inum, &parent,
 						    acceptable, context);
 }
 
@@ -102,9 +93,6 @@ static int gfs2_encode_fh(struct dentry 
 	fh[5] = cpu_to_be32(ip->i_no_formal_ino & 0xFFFFFFFF);
 	fh[6] = cpu_to_be32(ip->i_no_addr >> 32);
 	fh[7] = cpu_to_be32(ip->i_no_addr & 0xFFFFFFFF);
-
-	fh[8]  = cpu_to_be32(inode->i_mode);
-	fh[9]  = 0;	/* pad to double word */
 	*len = GFS2_LARGE_FH_SIZE;
 
 	iput(inode);
@@ -201,8 +189,7 @@ static struct dentry *gfs2_get_parent(st
 static struct dentry *gfs2_get_dentry(struct super_block *sb, void *inum_obj)
 {
 	struct gfs2_sbd *sdp = sb->s_fs_info;
-	struct gfs2_fh_obj *fh_obj = (struct gfs2_fh_obj *)inum_obj;
-	struct gfs2_inum_host *inum = &fh_obj->this;
+	struct gfs2_inum_host *inum = inum_obj;
 	struct gfs2_holder i_gh, ri_gh, rgd_gh;
 	struct gfs2_rgrpd *rgd;
 	struct inode *inode;
@@ -245,9 +232,9 @@ static struct dentry *gfs2_get_dentry(st
 	gfs2_glock_dq_uninit(&rgd_gh);
 	gfs2_glock_dq_uninit(&ri_gh);
 
-	inode = gfs2_inode_lookup(sb, fh_obj->imode,
+	inode = gfs2_inode_lookup(sb, DT_UNKNOWN,
 					inum->no_addr,
-					inum->no_formal_ino);
+					0);
 	if (!inode)
 		goto fail;
 	if (IS_ERR(inode)) {
@@ -260,6 +247,11 @@ static struct dentry *gfs2_get_dentry(st
 		iput(inode);
 		goto fail;
 	}
+
+	/* Pick up the works we bypass in gfs2_inode_lookup */
+	if (inode->i_state & I_NEW)
+		gfs2_set_iop(inode);
+
 	if (GFS2_I(inode)->i_no_formal_ino != inum->no_formal_ino) {
 		iput(inode);
 		goto fail;
--- gfs2-nfs2/fs/gfs2/rgrp.c	2007-06-27 01:18:09.000000000 -0400
+++ gfs2-nfs/fs/gfs2/rgrp.c	2007-06-27 14:55:11.000000000 -0400
@@ -873,7 +873,7 @@ static struct inode *try_rgrp_unlink(str
 			continue;
 		*last_unlinked = no_addr;
 		inode = gfs2_inode_lookup(rgd->rd_sbd->sd_vfs, DT_UNKNOWN,
-					no_addr, 0);
+					no_addr, -1);
 		if (!IS_ERR(inode))
 			return inode;
 	}