/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
* vim:expandtab:shiftwidth=8:tabstop=8:
*
- * linux/mds/mds_extN.c
- *
+ * lustre/mds/mds_extN.c
* Lustre Metadata Server (mds) journal abstraction routines
*
- * Copyright (C) 2002 Cluster File Systems, Inc.
- * author: Andreas Dilger <adilger@clusterfs.com>
+ * Copyright (C) 2002 Cluster File Systems, Inc.
+ * Author: Andreas Dilger <adilger@clusterfs.com>
+ *
+ * This file is part of Lustre, http://www.lustre.org.
*
- * This code is issued under the GNU General Public License.
- * See the file COPYING in this distribution
+ * Lustre is free software; you can redistribute it and/or
+ * modify it under the terms of version 2 of the GNU General Public
+ * License as published by the Free Software Foundation.
*
+ * Lustre is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with Lustre; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#define DEBUG_SUBSYSTEM S_MDS
#include <linux/fs.h>
#include <linux/jbd.h>
+#include <linux/slab.h>
#include <linux/extN_fs.h>
#include <linux/extN_jbd.h>
#include <linux/extN_xattr.h>
+#include <linux/kp30.h>
#include <linux/lustre_mds.h>
+#include <linux/obd.h>
#include <linux/module.h>
#include <linux/obd_lov.h>
static struct mds_fs_operations mds_extN_fs_ops;
-static kmem_cache_t *jcb_cache;
-static int jcb_cache_count;
+static kmem_cache_t *mcb_cache;
+static int mcb_cache_count;
struct mds_cb_data {
struct journal_callback cb_jcb;
#define EXTN_XATTR_INDEX_LUSTRE 5
#define XATTR_LUSTRE_MDS_OBJID "system.lustre_mds_objid"
-#define XATTR_MDS_MO_MAGIC 0x4711
+#define XATTR_MDS_MO_MAGIC 0xEA0BD047
/*
* We don't currently need any additional blocks for rmdir and
{
/* For updates to the last recieved file */
int nblocks = EXTN_DATA_TRANS_BLOCKS;
+ void *handle;
switch(op) {
case MDS_FSOP_RMDIR:
LBUG();
}
- return journal_start(EXTN_JOURNAL(inode), nblocks);
+ lock_kernel();
+ handle = journal_start(EXTN_JOURNAL(inode), nblocks);
+ unlock_kernel();
+
+ return handle;
}
static int mds_extN_commit(struct inode *inode, void *handle)
{
- return journal_stop((handle_t *)handle);
+ int rc;
+
+ lock_kernel();
+ rc = journal_stop((handle_t *)handle);
+ unlock_kernel();
+
+ return rc;
}
static int mds_extN_setattr(struct dentry *dentry, void *handle,
struct iattr *iattr)
{
struct inode *inode = dentry->d_inode;
+ int rc;
+ lock_kernel();
if (inode->i_op->setattr)
- return inode->i_op->setattr(dentry, iattr);
+ rc = inode->i_op->setattr(dentry, iattr);
else
- return inode_setattr(inode, iattr);
+ rc = inode_setattr(inode, iattr);
+
+ unlock_kernel();
+
+ return rc;
}
-static int mds_extN_set_obdo(struct inode *inode, void *handle,
- struct obdo *obdo)
+static int mds_extN_set_md(struct inode *inode, void *handle,
+ struct lov_mds_md *lmm)
{
- struct mds_objid *data;
int rc;
-
-
- lock_kernel();
down(&inode->i_sem);
- if (obdo == NULL)
- rc = extN_xattr_set(handle, inode, EXTN_XATTR_INDEX_LUSTRE,
- XATTR_LUSTRE_MDS_OBJID, NULL, 0, 0);
- else {
- data = (struct mds_objid *)obdo->o_inline;
- data->mo_magic = cpu_to_le64(XATTR_MDS_MO_MAGIC);
- rc = extN_xattr_set(handle, inode, EXTN_XATTR_INDEX_LUSTRE,
- XATTR_LUSTRE_MDS_OBJID, obdo->o_inline,
- OBD_INLINESZ, XATTR_CREATE);
- }
- up(&inode->i_sem);
+ lock_kernel();
+ rc = extN_xattr_set(handle, inode, EXTN_XATTR_INDEX_LUSTRE,
+ XATTR_LUSTRE_MDS_OBJID, lmm,
+ lmm ? lmm->lmm_easize : 0, 0);
unlock_kernel();
+ up(&inode->i_sem);
- if (rc)
- CERROR("error adding objectid %Ld to inode %ld\n",
- (unsigned long long)obdo->o_id, inode->i_ino);
+ if (rc) {
+ CERROR("error adding objectid "LPX64" to inode %ld: %d\n",
+ lmm->lmm_object_id, inode->i_ino, rc);
+ if (rc != -ENOSPC) LBUG();
+ }
return rc;
}
-static int mds_extN_get_obdo(struct inode *inode, struct obdo *obdo)
+static int mds_extN_get_md(struct inode *inode, struct lov_mds_md *lmm)
{
- struct mds_objid *data;
int rc;
+ int size = lmm->lmm_easize;
- lock_kernel();
down(&inode->i_sem);
+ lock_kernel();
rc = extN_xattr_get(inode, EXTN_XATTR_INDEX_LUSTRE,
- XATTR_LUSTRE_MDS_OBJID, obdo->o_inline,
- OBD_INLINESZ);
- data = (struct mds_objid *)obdo->o_inline;
-
- up(&inode->i_sem);
+ XATTR_LUSTRE_MDS_OBJID, lmm, size);
unlock_kernel();
+ up(&inode->i_sem);
+
+ /* This gives us the MD size */
+ if (lmm == NULL)
+ return rc;
if (rc < 0) {
CDEBUG(D_INFO, "error getting EA %s from MDS inode %ld: "
"rc = %d\n", XATTR_LUSTRE_MDS_OBJID, inode->i_ino, rc);
- obdo->o_id = 0;
- } else if (data->mo_magic != cpu_to_le64(XATTR_MDS_MO_MAGIC)) {
- CERROR("MDS object id %Ld has bad magic %Lx\n",
- (unsigned long long)obdo->o_id,
- (unsigned long long)le64_to_cpu(data->mo_magic));
- rc = -EINVAL;
- } else {
- /* This field is byteswapped because it appears in the
- * catalogue. All others are opaque to the MDS */
- obdo->o_id = le64_to_cpu(data->mo_lov_md.lmd_object_id);
- obdo->o_mode = S_IFREG;
- obdo->o_valid |= OBD_MD_FLID | OBD_MD_FLINLINE | OBD_MD_FLMODE;
+ memset(lmm, 0, size);
+ return rc;
}
-#warning FIXME: pass this buffer to caller for transmission when size exceeds OBD_INLINESZ
+ /* This field is byteswapped because it appears in the
+ * catalogue. All others are opaque to the MDS */
+ lmm->lmm_object_id = le64_to_cpu(lmm->lmm_object_id);
+
return rc;
}
EXIT;
return;
}
- if (mds_extN_set_obdo(inode, handle, NULL))
- CERROR("error clearing obdo on %ld\n", inode->i_ino);
+ if (mds_extN_set_md(inode, handle, NULL))
+ CERROR("error clearing objid on %ld\n", inode->i_ino);
if (mds_extN_fs_ops.cl_delete_inode)
mds_extN_fs_ops.cl_delete_inode(inode);
mds_extN_fs_ops.cl_delete_inode(inode);
}
-static void mds_extN_callback_status(void *jcb, int error)
+static void mds_extN_callback_status(struct journal_callback *jcb, int error)
{
struct mds_cb_data *mcb = (struct mds_cb_data *)jcb;
- CDEBUG(D_EXT2, "got callback for last_rcvd %Ld: rc = %d\n",
+ CDEBUG(D_EXT2, "got callback for last_rcvd "LPD64": rc = %d\n",
mcb->cb_last_rcvd, error);
if (!error && mcb->cb_last_rcvd > mcb->cb_mds->mds_last_committed)
mcb->cb_mds->mds_last_committed = mcb->cb_last_rcvd;
- kmem_cache_free(jcb_cache, jcb);
- --jcb_cache_count;
-}
-
-#ifdef HAVE_JOURNAL_CALLBACK
-static void mds_extN_callback_func(void *cb_data)
-{
- mds_extN_callback_status(cb_data, 0);
+ kmem_cache_free(mcb_cache, mcb);
+ --mcb_cache_count;
}
-#endif
static int mds_extN_set_last_rcvd(struct mds_obd *mds, void *handle)
{
struct mds_cb_data *mcb;
- mcb = kmem_cache_alloc(jcb_cache, GFP_NOFS);
+ mcb = kmem_cache_alloc(mcb_cache, GFP_NOFS);
if (!mcb)
RETURN(-ENOMEM);
- ++jcb_cache_count;
+ ++mcb_cache_count;
mcb->cb_mds = mds;
mcb->cb_last_rcvd = mds->mds_last_rcvd;
#ifdef HAVE_JOURNAL_CALLBACK_STATUS
- CDEBUG(D_EXT2, "set callback for last_rcvd: %Ld\n",
- (unsigned long long)mcb->cb_last_rcvd);
+ CDEBUG(D_EXT2, "set callback for last_rcvd: "LPD64"\n",
+ mcb->cb_last_rcvd);
+ lock_kernel();
+ /* Note that an "incompatible pointer" warning here is OK for now */
journal_callback_set(handle, mds_extN_callback_status,
- (void *)mcb);
-#elif defined(HAVE_JOURNAL_CALLBACK)
- /* XXX original patch version - remove soon */
-#warning "using old journal callback kernel patch, please update"
- CDEBUG(D_EXT2, "set callback for last_rcvd: %Ld\n",
- (unsigned long long)mcb->cb_last_rcvd);
- journal_callback_set(handle, mds_extN_callback_func, mcb);
+ (struct journal_callback *)mcb);
+ unlock_kernel();
#else
#warning "no journal callback kernel patch, faking it..."
{
CERROR("no journal callback kernel patch, faking it...\n");
next = jiffies + 300 * HZ;
}
- }
+
mds_extN_callback_status((struct journal_callback *)mcb, 0);
#endif
return 0;
}
+/*
+ * We need to hack the return value for the free inode counts because
+ * the current EA code requires one filesystem block per inode with EAs,
+ * so it is possible to run out of blocks before we run out of inodes.
+ *
+ * This can be removed when the extN EA code is fixed.
+ */
+static int mds_extN_statfs(struct super_block *sb, struct statfs *sfs)
+{
+ int rc = vfs_statfs(sb, sfs);
+
+ if (!rc && sfs->f_bfree < sfs->f_ffree)
+ sfs->f_ffree = sfs->f_bfree;
+
+ return rc;
+}
+
static struct mds_fs_operations mds_extN_fs_ops = {
+ fs_owner: THIS_MODULE,
fs_start: mds_extN_start,
fs_commit: mds_extN_commit,
fs_setattr: mds_extN_setattr,
- fs_set_obdo: mds_extN_set_obdo,
- fs_get_obdo: mds_extN_get_obdo,
+ fs_set_md: mds_extN_set_md,
+ fs_get_md: mds_extN_get_md,
fs_readpage: mds_extN_readpage,
fs_delete_inode: mds_extN_delete_inode,
cl_delete_inode: clear_inode,
fs_journal_data: mds_extN_journal_data,
fs_set_last_rcvd: mds_extN_set_last_rcvd,
+ fs_statfs: mds_extN_statfs,
};
static int __init mds_extN_init(void)
int rc;
//rc = extN_xattr_register();
- jcb_cache = kmem_cache_create("mds_extN_jcb",
+ mcb_cache = kmem_cache_create("mds_extN_mcb",
sizeof(struct mds_cb_data), 0,
0, NULL, NULL);
- if (!jcb_cache) {
+ if (!mcb_cache) {
CERROR("error allocating MDS journal callback cache\n");
GOTO(out, rc = -ENOMEM);
}
+
rc = mds_register_fs_type(&mds_extN_fs_ops, "extN");
if (rc)
- kmem_cache_destroy(jcb_cache);
+ kmem_cache_destroy(mcb_cache);
out:
return rc;
}
int rc;
mds_unregister_fs_type("extN");
- rc = kmem_cache_destroy(jcb_cache);
+ rc = kmem_cache_destroy(mcb_cache);
- if (rc || jcb_cache_count) {
+ if (rc || mcb_cache_count) {
CERROR("can't free MDS callback cache: count %d, rc = %d\n",
- jcb_cache_count, rc);
+ mcb_cache_count, rc);
}
//rc = extN_xattr_unregister();
}
-MODULE_AUTHOR("Cluster File Systems, Inc. <adilger@clusterfs.com>");
+MODULE_AUTHOR("Cluster File Systems, Inc. <info@clusterfs.com>");
MODULE_DESCRIPTION("Lustre MDS extN Filesystem Helper v0.1");
MODULE_LICENSE("GPL");