Whamcloud - gitweb
LU-1330 obdclass: add obd_target.h
[fs/lustre-release.git] / lustre / include / lustre_disk.h
index 6c46056..5081819 100644 (file)
@@ -1,6 +1,4 @@
-/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * vim:expandtab:shiftwidth=8:tabstop=8:
- *
+/*
  * GPL HEADER START
  *
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
@@ -28,6 +26,8 @@
 /*
  * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
  * Use is subject to license terms.
+ *
+ * Copyright (c) 2011, 2013, Intel Corporation.
  */
 /*
  * This file is part of Lustre, http://www.lustre.org/
 #define MOUNT_DATA_FILE    MOUNT_CONFIGS_DIR"/"CONFIGS_FILE
 #define LAST_RCVD         "last_rcvd"
 #define LOV_OBJID         "lov_objid"
+#define LOV_OBJSEQ             "lov_objseq"
 #define HEALTH_CHECK      "health_check"
 #define CAPA_KEYS         "capa_keys"
 #define CHANGELOG_USERS   "changelog_users"
-
+#define MGS_NIDTBL_DIR    "NIDTBL_VERSIONS"
+#define QMT_DIR           "quota_master"
+#define QSD_DIR           "quota_slave"
 
 /****************** persistent mount data *********************/
 
 #define LDD_F_SV_TYPE_MDT   0x0001
 #define LDD_F_SV_TYPE_OST   0x0002
 #define LDD_F_SV_TYPE_MGS   0x0004
+#define LDD_F_SV_TYPE_MASK (LDD_F_SV_TYPE_MDT  | \
+                            LDD_F_SV_TYPE_OST  | \
+                            LDD_F_SV_TYPE_MGS)
 #define LDD_F_SV_ALL        0x0008
 /** need an index assignment */
 #define LDD_F_NEED_INDEX    0x0010
 #define LDD_F_UPGRADE14     0x0200
 /** process as lctl conf_param */
 #define LDD_F_PARAM         0x0400
-/** backend fs make use of IAM directory format. */
-#define LDD_F_IAM_DIR       0x0800
+/** all nodes are specified as service nodes */
+#define LDD_F_NO_PRIMNODE   0x1000
+/** IR enable flag */
+#define LDD_F_IR_CAPABLE    0x2000
+/** the MGS refused to register the target. */
+#define LDD_F_ERROR         0x4000
+
+/* opc for target register */
+#define LDD_F_OPC_REG   0x10000000
+#define LDD_F_OPC_UNREG 0x20000000
+#define LDD_F_OPC_READY 0x40000000
+#define LDD_F_OPC_MASK  0xf0000000
+
+#define LDD_F_ONDISK_MASK  (LDD_F_SV_TYPE_MASK)
+
+#define LDD_F_MASK          0xFFFF
 
 enum ldd_mount_type {
-        LDD_MT_EXT3 = 0,
-        LDD_MT_LDISKFS,
-        LDD_MT_SMFS,
-        LDD_MT_REISERFS,
-        LDD_MT_LDISKFS2,
-        LDD_MT_LAST
+       LDD_MT_EXT3 = 0,
+       LDD_MT_LDISKFS,
+       LDD_MT_SMFS,
+       LDD_MT_REISERFS,
+       LDD_MT_LDISKFS2,
+       LDD_MT_ZFS,
+       LDD_MT_LAST
 };
 
 static inline char *mt_str(enum ldd_mount_type mt)
@@ -104,11 +125,25 @@ static inline char *mt_str(enum ldd_mount_type mt)
                 "ldiskfs",
                 "smfs",
                 "reiserfs",
-                "ldiskfs2"
+               "ldiskfs2",
+               "zfs",
         };
         return mount_type_string[mt];
 }
 
+static inline char *mt_type(enum ldd_mount_type mt)
+{
+       static char *mount_type_string[] = {
+               "osd-ldiskfs",
+               "osd-ldiskfs",
+               "osd-smfs",
+               "osd-reiserfs",
+               "osd-ldiskfs",
+               "osd-zfs",
+       };
+       return mount_type_string[mt];
+}
+
 #define LDD_INCOMPAT_SUPP 0
 #define LDD_ROCOMPAT_SUPP 0
 
@@ -137,10 +172,13 @@ struct lustre_disk_data {
 /*8192*/char       ldd_params[4096];     /* key=value pairs */
 };
 
-#define IS_MDT(data)   ((data)->ldd_flags & LDD_F_SV_TYPE_MDT)
-#define IS_OST(data)   ((data)->ldd_flags & LDD_F_SV_TYPE_OST)
-#define IS_MGS(data)  ((data)->ldd_flags & LDD_F_SV_TYPE_MGS)
-#define MT_STR(data)   mt_str((data)->ldd_mount_type)
+
+#define IS_MDT(data)    ((data)->lsi_flags & LDD_F_SV_TYPE_MDT)
+#define IS_OST(data)    ((data)->lsi_flags & LDD_F_SV_TYPE_OST)
+#define IS_MGS(data)    ((data)->lsi_flags & LDD_F_SV_TYPE_MGS)
+#define IS_SERVER(data) ((data)->lsi_flags & (LDD_F_SV_TYPE_MGS | \
+                        LDD_F_SV_TYPE_MDT | LDD_F_SV_TYPE_OST))
+#define MT_STR(data)    mt_str((data)->ldd_mount_type)
 
 /* Make the mdt/ost server obd name based on the filesystem name */
 static inline int server_make_name(__u32 flags, __u16 index, char *fs,
@@ -148,9 +186,11 @@ static inline int server_make_name(__u32 flags, __u16 index, char *fs,
 {
         if (flags & (LDD_F_SV_TYPE_MDT | LDD_F_SV_TYPE_OST)) {
                 if (!(flags & LDD_F_SV_ALL))
-                        sprintf(name, "%.8s-%s%04x", fs,
-                                (flags & LDD_F_SV_TYPE_MDT) ? "MDT" : "OST",
-                                index);
+                       sprintf(name, "%.8s%c%s%04x", fs,
+                               (flags & LDD_F_VIRGIN) ? ':' :
+                                       ((flags & LDD_F_WRITECONF) ? '=' : '-'),
+                               (flags & LDD_F_SV_TYPE_MDT) ? "MDT" : "OST",
+                               index);
         } else if (flags & LDD_F_SV_TYPE_MGS) {
                 sprintf(name, "MGS");
         } else {
@@ -160,10 +200,6 @@ static inline int server_make_name(__u32 flags, __u16 index, char *fs,
         return 0;
 }
 
-/* Get the index from the obd name */
-int server_name2index(char *svname, __u32 *idx, char **endptr);
-
-
 /****************** mount command *********************/
 
 /* The lmd is only used internally by Lustre; mount simply passes
@@ -184,7 +220,10 @@ struct lustre_mount_data {
         char      *lmd_mgssec;        /* sptlrpc flavor to mgs */
         char      *lmd_opts;          /* lustre mount options (as opposed to
                                          _device_ mount options) */
+       char      *lmd_params;        /* lustre params */
         __u32     *lmd_exclude;       /* array of OSTs to ignore */
+       char    *lmd_mgs;           /* MGS nid */
+       char    *lmd_osd_type;      /* OSD type */
 };
 
 #define LMD_FLG_SERVER       0x0001  /* Mounting a server */
@@ -195,6 +234,13 @@ struct lustre_mount_data {
 #define LMD_FLG_NOMGS        0x0020  /* Only start target for servers, reusing
                                         existing MGS services */
 #define LMD_FLG_WRITECONF    0x0040  /* Rewrite config log */
+#define LMD_FLG_NOIR         0x0080  /* NO imperative recovery */
+#define LMD_FLG_NOSCRUB             0x0100  /* Do not trigger scrub automatically */
+#define LMD_FLG_MGS         0x0200  /* Also start MGS along with server */
+#define LMD_FLG_IAM         0x0400  /* IAM dir */
+#define LMD_FLG_NO_PRIMNODE  0x0800  /* all nodes are service nodes */
+#define LMD_FLG_VIRGIN      0x1000  /* the service registers first time */
+#define LMD_FLG_UPDATE      0x2000  /* update parameters */
 
 #define lmd_is_client(x) ((x)->lmd_flags & LMD_FLG_CLIENT)
 
@@ -216,14 +262,14 @@ struct lustre_mount_data {
 
 /*
  * This limit is arbitrary (131072 clients on x86), but it is convenient to use
- * 2^n * CFS_PAGE_SIZE * 8 for the number of bits that fit an order-n allocation.
+ * 2^n * PAGE_CACHE_SIZE * 8 for the number of bits that fit an order-n allocation.
  * If we need more than 131072 clients (order-2 allocation on x86) then this
  * should become an array of single-page pointers that are allocated on demand.
  */
-#if (128 * 1024UL) > (CFS_PAGE_SIZE * 8)
+#if (128 * 1024UL) > (PAGE_CACHE_SIZE * 8)
 #define LR_MAX_CLIENTS (128 * 1024UL)
 #else
-#define LR_MAX_CLIENTS (CFS_PAGE_SIZE * 8)
+#define LR_MAX_CLIENTS (PAGE_CACHE_SIZE * 8)
 #endif
 
 /** COMPAT_146: this is an OST (temporary) */
@@ -252,7 +298,14 @@ struct lustre_mount_data {
 #define OBD_INCOMPAT_IAM_DIR    0x00000040
 /** LMA attribute contains per-inode incompatible flags */
 #define OBD_INCOMPAT_LMA        0x00000080
-
+/** lmm_stripe_count has been shrunk from __u32 to __u16 and the remaining 16
+ * bits are now used to store a generation. Once we start changing the layout
+ * and bumping the generation, old versions expecting a 32-bit lmm_stripe_count
+ * will be confused by interpreting stripe_count | gen << 16 as the actual
+ * stripe count */
+#define OBD_INCOMPAT_LMM_VER    0x00000100
+/** multiple OI files for MDT */
+#define OBD_INCOMPAT_MULTI_OI   0x00000200
 
 /* Data stored per server at the head of the last_rcvd file.  In le32 order.
    This should be common to filter_internal.h, lustre_mds.h */
@@ -271,8 +324,8 @@ struct lr_server_data {
         __u64 lsd_catalog_oid;     /* recovery catalog object id */
         __u32 lsd_catalog_ogen;    /* recovery catalog inode generation */
         __u8  lsd_peeruuid[40];    /* UUID of MDS associated with this OST */
-        __u32 lsd_ost_index;       /* index number of OST in LOV */
-        __u32 lsd_mdt_index;       /* index number of MDT in LMV */
+       __u32 lsd_osd_index;       /* index number of OST in LOV */
+       __u32 lsd_padding1;        /* was lsd_mdt_index, unused in 2.4.0 */
         __u32 lsd_start_epoch;     /* VBR: start epoch from last boot */
         /** transaction values since lsd_trans_table_time */
         __u64 lsd_trans_table[LR_EXPIRE_INTERVALS];
@@ -311,7 +364,7 @@ static inline void check_lcd(char *obd_name, int index,
                 lcd->lcd_uuid[length - 1] = '\0';
 
                 LCONSOLE_ERROR("the client UUID (%s) on %s for exports"
-                               "stored in last_rcvd(index = %d) is bad!\n", 
+                               "stored in last_rcvd(index = %d) is bad!\n",
                                lcd->lcd_uuid, obd_name, index);
         }
 }
@@ -320,55 +373,55 @@ static inline void check_lcd(char *obd_name, int index,
 static inline void lsd_le_to_cpu(struct lr_server_data *buf,
                                  struct lr_server_data *lsd)
 {
-        int i;
-        memcpy(lsd->lsd_uuid, buf->lsd_uuid, sizeof (lsd->lsd_uuid));
-        lsd->lsd_last_transno     = le64_to_cpu(buf->lsd_last_transno);
-        lsd->lsd_compat14         = le64_to_cpu(buf->lsd_compat14);
-        lsd->lsd_mount_count      = le64_to_cpu(buf->lsd_mount_count);
-        lsd->lsd_feature_compat   = le32_to_cpu(buf->lsd_feature_compat);
-        lsd->lsd_feature_rocompat = le32_to_cpu(buf->lsd_feature_rocompat);
-        lsd->lsd_feature_incompat = le32_to_cpu(buf->lsd_feature_incompat);
-        lsd->lsd_server_size      = le32_to_cpu(buf->lsd_server_size);
-        lsd->lsd_client_start     = le32_to_cpu(buf->lsd_client_start);
-        lsd->lsd_client_size      = le16_to_cpu(buf->lsd_client_size);
-        lsd->lsd_subdir_count     = le16_to_cpu(buf->lsd_subdir_count);
-        lsd->lsd_catalog_oid      = le64_to_cpu(buf->lsd_catalog_oid);
-        lsd->lsd_catalog_ogen     = le32_to_cpu(buf->lsd_catalog_ogen);
-        memcpy(lsd->lsd_peeruuid, buf->lsd_peeruuid, sizeof(lsd->lsd_peeruuid));
-        lsd->lsd_ost_index        = le32_to_cpu(buf->lsd_ost_index);
-        lsd->lsd_mdt_index        = le32_to_cpu(buf->lsd_mdt_index);
-        lsd->lsd_start_epoch      = le32_to_cpu(buf->lsd_start_epoch);
-        for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
-                lsd->lsd_trans_table[i] = le64_to_cpu(buf->lsd_trans_table[i]);
-        lsd->lsd_trans_table_time = le32_to_cpu(buf->lsd_trans_table_time);
-        lsd->lsd_expire_intervals = le32_to_cpu(buf->lsd_expire_intervals);
+       int i;
+       memcpy(lsd->lsd_uuid, buf->lsd_uuid, sizeof(lsd->lsd_uuid));
+       lsd->lsd_last_transno     = le64_to_cpu(buf->lsd_last_transno);
+       lsd->lsd_compat14         = le64_to_cpu(buf->lsd_compat14);
+       lsd->lsd_mount_count      = le64_to_cpu(buf->lsd_mount_count);
+       lsd->lsd_feature_compat   = le32_to_cpu(buf->lsd_feature_compat);
+       lsd->lsd_feature_rocompat = le32_to_cpu(buf->lsd_feature_rocompat);
+       lsd->lsd_feature_incompat = le32_to_cpu(buf->lsd_feature_incompat);
+       lsd->lsd_server_size      = le32_to_cpu(buf->lsd_server_size);
+       lsd->lsd_client_start     = le32_to_cpu(buf->lsd_client_start);
+       lsd->lsd_client_size      = le16_to_cpu(buf->lsd_client_size);
+       lsd->lsd_subdir_count     = le16_to_cpu(buf->lsd_subdir_count);
+       lsd->lsd_catalog_oid      = le64_to_cpu(buf->lsd_catalog_oid);
+       lsd->lsd_catalog_ogen     = le32_to_cpu(buf->lsd_catalog_ogen);
+       memcpy(lsd->lsd_peeruuid, buf->lsd_peeruuid, sizeof(lsd->lsd_peeruuid));
+       lsd->lsd_osd_index        = le32_to_cpu(buf->lsd_osd_index);
+       lsd->lsd_padding1        = le32_to_cpu(buf->lsd_padding1);
+       lsd->lsd_start_epoch      = le32_to_cpu(buf->lsd_start_epoch);
+       for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
+               lsd->lsd_trans_table[i] = le64_to_cpu(buf->lsd_trans_table[i]);
+       lsd->lsd_trans_table_time = le32_to_cpu(buf->lsd_trans_table_time);
+       lsd->lsd_expire_intervals = le32_to_cpu(buf->lsd_expire_intervals);
 }
 
 static inline void lsd_cpu_to_le(struct lr_server_data *lsd,
                                  struct lr_server_data *buf)
 {
-        int i;
-        memcpy(buf->lsd_uuid, lsd->lsd_uuid, sizeof (buf->lsd_uuid));
-        buf->lsd_last_transno     = cpu_to_le64(lsd->lsd_last_transno);
-        buf->lsd_compat14         = cpu_to_le64(lsd->lsd_compat14);
-        buf->lsd_mount_count      = cpu_to_le64(lsd->lsd_mount_count);
-        buf->lsd_feature_compat   = cpu_to_le32(lsd->lsd_feature_compat);
-        buf->lsd_feature_rocompat = cpu_to_le32(lsd->lsd_feature_rocompat);
-        buf->lsd_feature_incompat = cpu_to_le32(lsd->lsd_feature_incompat);
-        buf->lsd_server_size      = cpu_to_le32(lsd->lsd_server_size);
-        buf->lsd_client_start     = cpu_to_le32(lsd->lsd_client_start);
-        buf->lsd_client_size      = cpu_to_le16(lsd->lsd_client_size);
-        buf->lsd_subdir_count     = cpu_to_le16(lsd->lsd_subdir_count);
-        buf->lsd_catalog_oid      = cpu_to_le64(lsd->lsd_catalog_oid);
-        buf->lsd_catalog_ogen     = cpu_to_le32(lsd->lsd_catalog_ogen);
-        memcpy(buf->lsd_peeruuid, lsd->lsd_peeruuid, sizeof(buf->lsd_peeruuid));
-        buf->lsd_ost_index        = cpu_to_le32(lsd->lsd_ost_index);
-        buf->lsd_mdt_index        = cpu_to_le32(lsd->lsd_mdt_index);
-        buf->lsd_start_epoch      = cpu_to_le32(lsd->lsd_start_epoch);
-        for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
-                buf->lsd_trans_table[i] = cpu_to_le64(lsd->lsd_trans_table[i]);
-        buf->lsd_trans_table_time = cpu_to_le32(lsd->lsd_trans_table_time);
-        buf->lsd_expire_intervals = cpu_to_le32(lsd->lsd_expire_intervals);
+       int i;
+       memcpy(buf->lsd_uuid, lsd->lsd_uuid, sizeof(buf->lsd_uuid));
+       buf->lsd_last_transno     = cpu_to_le64(lsd->lsd_last_transno);
+       buf->lsd_compat14         = cpu_to_le64(lsd->lsd_compat14);
+       buf->lsd_mount_count      = cpu_to_le64(lsd->lsd_mount_count);
+       buf->lsd_feature_compat   = cpu_to_le32(lsd->lsd_feature_compat);
+       buf->lsd_feature_rocompat = cpu_to_le32(lsd->lsd_feature_rocompat);
+       buf->lsd_feature_incompat = cpu_to_le32(lsd->lsd_feature_incompat);
+       buf->lsd_server_size      = cpu_to_le32(lsd->lsd_server_size);
+       buf->lsd_client_start     = cpu_to_le32(lsd->lsd_client_start);
+       buf->lsd_client_size      = cpu_to_le16(lsd->lsd_client_size);
+       buf->lsd_subdir_count     = cpu_to_le16(lsd->lsd_subdir_count);
+       buf->lsd_catalog_oid      = cpu_to_le64(lsd->lsd_catalog_oid);
+       buf->lsd_catalog_ogen     = cpu_to_le32(lsd->lsd_catalog_ogen);
+       memcpy(buf->lsd_peeruuid, lsd->lsd_peeruuid, sizeof(buf->lsd_peeruuid));
+       buf->lsd_osd_index        = cpu_to_le32(lsd->lsd_osd_index);
+       buf->lsd_padding1         = cpu_to_le32(lsd->lsd_padding1);
+       buf->lsd_start_epoch      = cpu_to_le32(lsd->lsd_start_epoch);
+       for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
+               buf->lsd_trans_table[i] = cpu_to_le64(lsd->lsd_trans_table[i]);
+       buf->lsd_trans_table_time = cpu_to_le32(lsd->lsd_trans_table_time);
+       buf->lsd_expire_intervals = cpu_to_le32(lsd->lsd_expire_intervals);
 }
 
 static inline void lcd_le_to_cpu(struct lsd_client_data *buf,
@@ -432,22 +485,29 @@ struct lustre_sb_info {
         int                       lsi_flags;
         struct obd_device        *lsi_mgc;     /* mgc obd */
         struct lustre_mount_data *lsi_lmd;     /* mount command info */
-        struct lustre_disk_data  *lsi_ldd;     /* mount info on-disk */
         struct ll_sb_info        *lsi_llsbi;   /* add'l client sbi info */
+       struct dt_device         *lsi_dt_dev;  /* dt device to access disk fs*/
         struct vfsmount          *lsi_srv_mnt; /* the one server mount */
         cfs_atomic_t              lsi_mounts;  /* references to the srv_mnt */
-        struct backing_dev_info   lsi_bdi;     /* each client mountpoint needs own backing_dev_info */
+       char                      lsi_svname[MTI_NAME_MAXLEN];
+       char                      lsi_osd_obdname[64];
+       char                      lsi_osd_uuid[64];
+       struct obd_export        *lsi_osd_exp;
+       char                      lsi_osd_type[16];
+       char                      lsi_fstype[16];
+        struct backing_dev_info   lsi_bdi;     /* each client mountpoint needs
+                                                  own backing_dev_info */
 };
 
-#define LSI_SERVER                       0x00000001
-#define LSI_UMOUNT_FORCE                 0x00000010
-#define LSI_UMOUNT_FAILOVER              0x00000020
-#define LSI_BDI_INITIALIZED              0x00000040
+#define LSI_UMOUNT_FAILOVER              0x00200000
+#define LSI_BDI_INITIALIZED              0x00400000
 
 #define     s2lsi(sb)        ((struct lustre_sb_info *)((sb)->s_fs_info))
 #define     s2lsi_nocast(sb) ((sb)->s_fs_info)
 
 #define     get_profile_name(sb)   (s2lsi(sb)->lsi_lmd->lmd_profile)
+#define            get_mount_flags(sb)    (s2lsi(sb)->lsi_lmd->lmd_flags)
+#define            get_mntdev_name(sb)    (s2lsi(sb)->lsi_lmd->lmd_dev)
 
 #endif /* __KERNEL__ */
 
@@ -463,26 +523,35 @@ struct lustre_mount_info {
 /****************** prototypes *********************/
 
 #ifdef __KERNEL__
-
 /* obd_mount.c */
-void lustre_register_client_fill_super(int (*cfs)(struct super_block *sb));
+int server_name2fsname(const char *svname, char *fsname, const char **endptr);
+int server_name2index(const char *svname, __u32 *idx, const char **endptr);
+int server_name2svname(const char *label, char *svname, const char **endptr,
+                      size_t svsize);
+
+int lustre_put_lsi(struct super_block *sb);
+int lustre_start_simple(char *obdname, char *type, char *uuid,
+                       char *s1, char *s2, char *s3, char *s4);
+int lustre_start_mgc(struct super_block *sb);
+void lustre_register_client_fill_super(int (*cfs)(struct super_block *sb,
+                                                 struct vfsmount *mnt));
 void lustre_register_kill_super_cb(void (*cfs)(struct super_block *sb));
-
-
 int lustre_common_put_super(struct super_block *sb);
-struct lustre_mount_info *server_find_mount_locked(const char *name);
+
+# ifdef HAVE_SERVER_SUPPORT
+/* obd_mount_server.c */
+int server_fill_super(struct super_block *sb);
 struct lustre_mount_info *server_get_mount(const char *name);
 struct lustre_mount_info *server_get_mount_2(const char *name);
 int server_put_mount(const char *name, struct vfsmount *mnt);
 int server_put_mount_2(const char *name, struct vfsmount *mnt);
-int server_register_target(struct super_block *sb);
 struct mgs_target_info;
-int server_mti_print(char *title, struct mgs_target_info *mti);
-
-/* mgc_request.c */
-int mgc_fsname2resid(char *fsname, struct ldlm_res_id *res_id);
+int server_mti_print(const char *title, struct mgs_target_info *mti);
+void server_calc_timeout(struct lustre_sb_info *lsi, struct obd_device *obd);
+# endif
 
-#endif
+int mgc_fsname2resid(char *fsname, struct ldlm_res_id *res_id, int type);
+#endif /* __KERNEL__ */
 
 /** @} disk */