[PATCH 1/4] pnfsd-exofs: Add autologin support to exofs

[Boaz Harrosh <bharrosh@panasas.com>]

Introduce sysfs infrastructure for exofs cluster filesystem.

Each OSD target shows up as below in the sysfs hierarchy:
	/sys/fs/exofs/<osdname>_<partition_id>/devX

where in devX 0 <= X < device_table_size. They are ordered
in device-table order as specified to the mkfs.exofs command

Each OSD device  devX has following attributes :
	osdname - ReadOnly
	systemid - ReadOnly
	uri - Read/Write

It is up to user-mode to update devX/uri for support of
autologin.

Also fixed is the xdr encoding of device_info to support
uri encoding.

Signed-off-by: Sachin Bhamare <sbhamare@panasas.com>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
---
 fs/exofs/Kbuild                |    2 +-
 fs/exofs/exofs.h               |    4 +
 fs/exofs/export.c              |    8 ++-
 fs/exofs/super.c               |   14 +++
 fs/exofs/sys.c                 |  202 ++++++++++++++++++++++++++++++++++++++++
 fs/exofs/sys.h                 |   41 ++++++++
 fs/exportfs/pnfs_osd_xdr_srv.c |   45 ++++++++-
 include/linux/pnfs_osd_xdr.h   |    5 +
 8 files changed, 316 insertions(+), 5 deletions(-)
 create mode 100644 fs/exofs/sys.c
 create mode 100644 fs/exofs/sys.h

diff --git a/fs/exofs/Kbuild b/fs/exofs/Kbuild
index 1118068..5f6f1eb 100644
--- a/fs/exofs/Kbuild
+++ b/fs/exofs/Kbuild
@@ -16,6 +16,6 @@
 libore-y := ore.o ore_raid.o
 obj-$(CONFIG_ORE) += libore.o
 
-exofs-y := inode.o file.o symlink.o namei.o dir.o super.o
+exofs-y := inode.o file.o symlink.o namei.o dir.o super.o sys.o
 exofs-$(CONFIG_PNFSD) +=  export.o
 obj-$(CONFIG_EXOFS_FS) += exofs.o
diff --git a/fs/exofs/exofs.h b/fs/exofs/exofs.h
index 94047b9..b298c7a 100644
--- a/fs/exofs/exofs.h
+++ b/fs/exofs/exofs.h
@@ -56,6 +56,9 @@
 struct exofs_dev {
 	struct ore_dev ored;
 	unsigned did;
+	unsigned urilen;
+	uint8_t *uri;
+	struct kobject ed_kobj;
 };
 /*
  * our extension to the in-memory superblock
@@ -73,6 +76,7 @@ struct exofs_sb_info {
 	struct ore_layout	layout;		/* Default files layout       */
 	struct ore_comp one_comp;		/* id & cred of partition id=0*/
 	struct ore_components oc;		/* comps for the partition    */
+	struct kobject	s_kobj;			/* holds per-sbi kobject      */
 };
 
 /*
diff --git a/fs/exofs/export.c b/fs/exofs/export.c
index 621bd11..a53f575 100644
--- a/fs/exofs/export.c
+++ b/fs/exofs/export.c
@@ -321,6 +321,7 @@ int exofs_get_device_info(struct super_block *sb, struct exp_xdr_stream *xdr,
 {
 	struct exofs_sb_info *sbi = sb->s_fs_info;
 	struct pnfs_osd_deviceaddr devaddr;
+	struct exofs_dev *edev;
 	const struct osd_dev_info *odi;
 	u64 devno = devid->devid;
 	__be32 *start;
@@ -334,7 +335,8 @@ int exofs_get_device_info(struct super_block *sb, struct exp_xdr_stream *xdr,
 		return -ENODEV;
 	}
 
-	odi = osduld_device_info(sbi->oc.ods[devno]->od);
+	edev = container_of(sbi->oc.ods[devno], typeof(*edev), ored);
+	odi = osduld_device_info(edev->ored.od);
 
 	devaddr.oda_systemid.len = odi->systemid_len;
 	devaddr.oda_systemid.data = (void *)odi->systemid; /* !const cast */
@@ -342,6 +344,10 @@ int exofs_get_device_info(struct super_block *sb, struct exp_xdr_stream *xdr,
 	devaddr.oda_osdname.len = odi->osdname_len ;
 	devaddr.oda_osdname.data = (void *)odi->osdname;/* !const cast */
 
+	devaddr.oda_targetaddr.ota_available = OBJ_OTA_AVAILABLE;
+	devaddr.oda_targetaddr.ota_netaddr.r_addr.data = (void *)edev->uri;
+	devaddr.oda_targetaddr.ota_netaddr.r_addr.len = edev->urilen;
+
 	/* skip opaque size, will be filled-in later */
 	start = exp_xdr_reserve_qwords(xdr, 1);
 	if (!start) {
diff --git a/fs/exofs/super.c b/fs/exofs/super.c
index 3c0d83a..a4834f7 100644
--- a/fs/exofs/super.c
+++ b/fs/exofs/super.c
@@ -39,6 +39,7 @@
 #include <linux/exportfs.h>
 #include <linux/slab.h>
 
+#include "sys.h"
 #include "exofs.h"
 
 #define EXOFS_DBGMSG2(M...) do {} while (0)
@@ -472,6 +473,7 @@ static void exofs_put_super(struct super_block *sb)
 	_exofs_print_device("Unmounting", NULL, ore_comp_dev(&sbi->oc, 0),
 			    sbi->one_comp.obj.partition);
 
+	exofs_sysfs_cluster_del(sbi);
 	bdi_destroy(&sbi->bdi);
 	exofs_free_sbi(sbi);
 	sb->s_fs_info = NULL;
@@ -632,6 +634,11 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info *sbi,
 	memcpy(&sbi->oc.ods[numdevs], &sbi->oc.ods[0],
 		(numdevs - 1) * sizeof(sbi->oc.ods[0]));
 
+	/* create sysfs entries to hold the current exofs cluster instance.
+	 * There will be one sysfs dirent for cluster osdname per exofs mount
+	 */
+	exofs_sysfs_cluster_add(&dt->dt_dev_table[0], sbi);
+
 	for (i = 0; i < numdevs; i++) {
 		struct exofs_fscb fscb;
 		struct osd_dev_info odi;
@@ -657,6 +664,7 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info *sbi,
 			eds[i].ored.od = fscb_od;
 			++sbi->oc.numdevs;
 			fscb_od = NULL;
+			exofs_sysfs_odev_add(&eds[i], sbi);
 			continue;
 		}
 
@@ -682,6 +690,7 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info *sbi,
 				  odi.osdname);
 			goto out;
 		}
+		exofs_sysfs_odev_add(&eds[i], sbi);
 
 		/* TODO: verify other information is correct and FS-uuid
 		 *	 matches. Benny what did you say about device table
@@ -845,6 +854,7 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
 		goto free_sbi;
 	}
 
+	exofs_sysfs_print();
 	_exofs_print_device("Mounting", opts->dev_name,
 			    ore_comp_dev(&sbi->oc, 0),
 			    sbi->one_comp.obj.partition);
@@ -1024,6 +1034,9 @@ static int __init init_exofs(void)
 	if (err)
 		goto out_d;
 
+	/* We don't fail if sysfs creation failed */
+	exofs_sysfs_init();
+
 	return 0;
 out_d:
 	destroy_inodecache();
@@ -1033,6 +1046,7 @@ out:
 
 static void __exit exit_exofs(void)
 {
+	exofs_sysfs_uninit();
 	unregister_filesystem(&exofs_type);
 	destroy_inodecache();
 }
diff --git a/fs/exofs/sys.c b/fs/exofs/sys.c
new file mode 100644
index 0000000..17c9c07
--- /dev/null
+++ b/fs/exofs/sys.c
@@ -0,0 +1,202 @@
+/*
+ * Copyright (C) 2012
+ * Sachin Bhamare <sbhamare@panasas.com>
+ *
+ * This file is part of exofs.
+ *
+ * exofs is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation.  Since it is based on ext2, and the only
+ * valid version of GPL for the Linux kernel is version 2, the only valid
+ * version of GPL for exofs is version 2.
+ *
+ * exofs 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 exofs; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+
+#include <linux/kobject.h>
+#include <linux/device.h>
+
+#include "sys.h"
+#include "exofs.h"
+
+struct odev_attr {
+	struct attribute attr;
+	ssize_t (*show)(struct exofs_dev *, char *);
+	ssize_t (*store)(struct exofs_dev *, const char *, size_t);
+};
+
+static ssize_t odev_attr_show(struct kobject *kobj, struct attribute *attr,
+		char *buf)
+{
+	struct exofs_dev *edp = container_of(kobj, struct exofs_dev, ed_kobj);
+	struct odev_attr *a = container_of(attr, struct odev_attr, attr);
+
+	return a->show ? a->show(edp, buf) : 0;
+}
+
+static ssize_t odev_attr_store(struct kobject *kobj, struct attribute *attr,
+		const char *buf, size_t len)
+{
+	struct exofs_dev *edp = container_of(kobj, struct exofs_dev, ed_kobj);
+	struct odev_attr *a = container_of(attr, struct odev_attr, attr);
+
+	return a->store ? a->store(edp, buf, len) : len;
+}
+
+static const struct sysfs_ops odev_attr_ops = {
+	.show  = odev_attr_show,
+	.store = odev_attr_store,
+};
+
+
+static struct kset *exofs_kset;
+
+static ssize_t osdname_show(struct exofs_dev *edp, char *buf)
+{
+	struct osd_dev *odev = edp->ored.od;
+	const struct osd_dev_info *odi = osduld_device_info(odev);
+
+	return snprintf(buf, odi->osdname_len + 1, "%s", odi->osdname);
+}
+
+static ssize_t systemid_show(struct exofs_dev *edp, char *buf)
+{
+	struct osd_dev *odev = edp->ored.od;
+	const struct osd_dev_info *odi = osduld_device_info(odev);
+
+	memcpy(buf, odi->systemid, odi->systemid_len);
+	return odi->systemid_len;
+}
+
+static ssize_t uri_show(struct exofs_dev *edp, char *buf)
+{
+	return snprintf(buf, edp->urilen, "%s", edp->uri);
+}
+
+static ssize_t uri_store(struct exofs_dev *edp, const char *buf, size_t len)
+{
+	edp->urilen = strlen(buf) + 1;
+	edp->uri = krealloc(edp->uri, edp->urilen, GFP_KERNEL);
+	strncpy(edp->uri, buf, edp->urilen);
+	return edp->urilen;
+}
+
+#define OSD_ATTR(name, mode, show, store) \
+	static struct odev_attr odev_attr_##name = \
+					__ATTR(name, mode, show, store)
+
+OSD_ATTR(osdname, S_IRUGO, osdname_show, NULL);
+OSD_ATTR(systemid, S_IRUGO, systemid_show, NULL);
+OSD_ATTR(uri, S_IRWXU, uri_show, uri_store);
+
+static struct attribute *odev_attrs[] = {
+	&odev_attr_osdname.attr,
+	&odev_attr_systemid.attr,
+	&odev_attr_uri.attr,
+	NULL,
+};
+
+static struct kobj_type odev_ktype = {
+	.default_attrs	= odev_attrs,
+	.sysfs_ops	= &odev_attr_ops,
+};
+
+static struct kobj_type uuid_ktype = {
+};
+
+void exofs_sysfs_print()
+{
+#ifdef CONFIG_EXOFS_DEBUG
+	struct kobject *k_name, *k_tmp;
+
+	list_for_each_entry_safe(k_name, k_tmp, &exofs_kset->list, entry) {
+		printk(KERN_INFO "%s: name %s ref %d\n",
+			__func__, kobject_name(k_name),
+			(int)atomic_read(&k_name->kref.refcount));
+	}
+#endif
+}
+/*
+ * This function removes all kobjects under exofs_kset
+ * At the end of it, exofs_kset kobject will have a refcount
+ * of 1 which gets decremented only on exofs module unload
+ */
+void exofs_sysfs_cluster_del(struct exofs_sb_info *sbi)
+{
+	struct kobject *k_name, *k_tmp;
+	struct kobject *s_kobj = &sbi->s_kobj;
+
+	list_for_each_entry_safe(k_name, k_tmp, &exofs_kset->list, entry) {
+		/* Remove all that are children of this SBI */
+		if (k_name->parent == s_kobj)
+			kobject_put(k_name);
+	}
+	kobject_put(s_kobj);
+}
+
+/*
+ * This function creates sysfs entries to hold the current exofs cluster
+ * instance (uniquely identified by osdname,pid tuple).
+ * This function gets called once per exofs mount instance.
+ */
+int exofs_sysfs_cluster_add(struct exofs_dt_device_info *dt_dev,
+		struct exofs_sb_info *sbi)
+{
+	struct kobject *s_kobj;
+	int retval = 0;
+	uint64_t pid = sbi->one_comp.obj.partition;
+
+	/* allocate new uuid dirent */
+	s_kobj = &sbi->s_kobj;
+	s_kobj->kset = exofs_kset;
+	retval = kobject_init_and_add(s_kobj, &uuid_ktype,
+			&exofs_kset->kobj,  "%s_%llx", dt_dev->osdname, pid);
+	if (retval) {
+		EXOFS_ERR("ERROR: Failed to create sysfs entry for "
+			"uuid-%s_%llx => %d\n", dt_dev->osdname, pid, retval);
+		return -ENOMEM;
+	}
+	return 0;
+}
+
+int exofs_sysfs_odev_add(struct exofs_dev *edev, struct exofs_sb_info *sbi)
+{
+	struct kobject *d_kobj;
+	int retval = 0;
+
+	/* create osd device group which contains following attributes
+	 * osdname, systemid & uri
+	 */
+	d_kobj = &edev->ed_kobj;
+	d_kobj->kset = exofs_kset;
+	retval = kobject_init_and_add(d_kobj, &odev_ktype,
+			&sbi->s_kobj, "dev%u", edev->did);
+	if (retval) {
+		EXOFS_ERR("ERROR: Failed to create sysfs entry for"
+				" device dev%u\n", edev->did);
+		return retval;
+	}
+	return 0;
+}
+
+int exofs_sysfs_init(void)
+{
+	exofs_kset = kset_create_and_add("exofs", NULL, fs_kobj);
+	if (!exofs_kset) {
+		EXOFS_ERR("ERROR: kset_create_and_add exofs failed\n");
+		return -ENOMEM;
+	}
+	return 0;
+}
+
+void exofs_sysfs_uninit(void)
+{
+	kset_unregister(exofs_kset);
+}
diff --git a/fs/exofs/sys.h b/fs/exofs/sys.h
new file mode 100644
index 0000000..cfabc1d
--- /dev/null
+++ b/fs/exofs/sys.h
@@ -0,0 +1,41 @@
+/*
+ * Copyright (C) 2012
+ * Sachin Bhamare <sbhamare@panasas.com>
+ *
+ * This file is part of exofs.
+ *
+ * exofs is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation.  Since it is based on ext2, and the only
+ * valid version of GPL for the Linux kernel is version 2, the only valid
+ * version of GPL for exofs is version 2.
+ *
+ * exofs 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 exofs; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+
+#ifndef __SYS_DOT_H__
+#define __SYS_DOT_H__
+
+struct exofs_dev;
+struct exofs_sb_info;
+struct exofs_dt_device_info;
+
+int exofs_sysfs_odev_add(struct exofs_dev *edev,
+			 struct exofs_sb_info *sbi);
+int exofs_sysfs_cluster_add(struct exofs_dt_device_info *dt_dev,
+		struct exofs_sb_info *sbi);
+void exofs_sysfs_cluster_del(struct exofs_sb_info *sbi);
+
+void exofs_sysfs_print(void);
+int exofs_sysfs_init(void);
+void exofs_sysfs_uninit(void);
+
+#endif /* __SYS_DOT_H__ */
+
diff --git a/fs/exportfs/pnfs_osd_xdr_srv.c b/fs/exportfs/pnfs_osd_xdr_srv.c
index 12a3bda..d0795c44 100644
--- a/fs/exportfs/pnfs_osd_xdr_srv.c
+++ b/fs/exportfs/pnfs_osd_xdr_srv.c
@@ -179,6 +179,42 @@ static enum nfsstat4 _encode_string(struct exp_xdr_stream *xdr,
 	return 0;
 }
 
+/* struct pnfs_osd_targetaddr {
+ *	u32				ota_available;
+ *	struct pnfs_osd_net_addr	ota_netaddr;
+ * };
+ */
+static inline enum nfsstat4 pnfs_osd_xdr_encode_targetaddr(
+	struct exp_xdr_stream *xdr,
+	struct pnfs_osd_targetaddr *taddr)
+{
+	__be32 *p;
+
+	/* ota_available */
+	p = exp_xdr_reserve_space(xdr, 4);
+	if (!p)
+		return NFS4ERR_TOOSMALL;
+	p = exp_xdr_encode_u32(p, taddr->ota_available);
+
+	/* encode r_netid */
+	p = exp_xdr_reserve_space(xdr, 4 + taddr->ota_netaddr.r_netid.len);
+	if (!p)
+		return NFS4ERR_TOOSMALL;
+
+	p = exp_xdr_encode_opaque(p,
+				taddr->ota_netaddr.r_netid.data,
+				taddr->ota_netaddr.r_netid.len);
+
+	/* encode r_addr */
+	p = exp_xdr_reserve_space(xdr, 4 + taddr->ota_netaddr.r_addr.len);
+	if (!p)
+		return NFS4ERR_TOOSMALL;
+	p = exp_xdr_encode_opaque(p,
+				taddr->ota_netaddr.r_addr.data,
+				taddr->ota_netaddr.r_addr.len);
+	return 0;
+}
+
 /* struct pnfs_osd_deviceaddr {
  *	struct pnfs_osd_targetid	oda_targetid;
  *	struct pnfs_osd_targetaddr	oda_targetaddr;
@@ -194,17 +230,20 @@ enum nfsstat4 pnfs_osd_xdr_encode_deviceaddr(
 	__be32 *p;
 	enum nfsstat4 err;
 
-	p = exp_xdr_reserve_space(xdr, 4 + 4 + sizeof(devaddr->oda_lun));
+	p = exp_xdr_reserve_space(xdr, sizeof(u32));
 	if (!p)
 		return NFS4ERR_TOOSMALL;
 
 	/* Empty oda_targetid */
 	p = exp_xdr_encode_u32(p, OBJ_TARGET_ANON);
 
-	/* Empty oda_targetaddr for now */
-	p = exp_xdr_encode_u32(p, 0);
+	/* oda_targetaddr */
+	err = pnfs_osd_xdr_encode_targetaddr(xdr, &devaddr->oda_targetaddr);
+	if (err)
+		return err;
 
 	/* oda_lun */
+	p = exp_xdr_reserve_space(xdr, sizeof(devaddr->oda_lun));
 	exp_xdr_encode_bytes(p, devaddr->oda_lun, sizeof(devaddr->oda_lun));
 
 	err = _encode_string(xdr, &devaddr->oda_systemid);
diff --git a/include/linux/pnfs_osd_xdr.h b/include/linux/pnfs_osd_xdr.h
index e0557c1..eeedf73 100644
--- a/include/linux/pnfs_osd_xdr.h
+++ b/include/linux/pnfs_osd_xdr.h
@@ -154,6 +154,11 @@ enum pnfs_osd_targetid_type {
 	OBJ_TARGET_SCSI_DEVICE_ID = 3,
 };
 
+enum pnfs_osd_target_ota {
+	OBJ_OTA_UNAVAILABLE = 0,
+	OBJ_OTA_AVAILABLE = 1,
+};
+
 /*   union pnfs_osd_targetid4 switch (pnfs_osd_targetid_type4 oti_type) {
  *       case OBJ_TARGET_SCSI_NAME:
  *           string              oti_scsi_name<>;
-- 
1.7.6.2