X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fllite%2Fllite_lib.c;h=4e49985e3ebdbaa944434ded9476a87bd0878b96;hp=cd336074c39b8c3fd5d7a50407ec62bdf5ee54cb;hb=322968acf183ab16d952cd3026f6580957b31259;hpb=96a5daa0c08d7b42ec368080a2a7f0dfb110ef98 diff --git a/lustre/llite/llite_lib.c b/lustre/llite/llite_lib.c index cd33607..4e49985 100644 --- a/lustre/llite/llite_lib.c +++ b/lustre/llite/llite_lib.c @@ -30,6 +30,9 @@ * Use is subject to license terms. */ /* + * Copyright (c) 2011 Whamcloud, Inc. + */ +/* * This file is part of Lustre, http://www.lustre.org/ * Lustre is a trademark of Sun Microsystems, Inc. * @@ -137,8 +140,10 @@ static struct ll_sb_info *ll_init_sbi(void) /* metadata statahead is enabled by default */ sbi->ll_sa_max = LL_SA_RPC_DEF; - atomic_set(&sbi->ll_sa_total, 0); - atomic_set(&sbi->ll_sa_wrong, 0); + cfs_atomic_set(&sbi->ll_sa_total, 0); + cfs_atomic_set(&sbi->ll_sa_wrong, 0); + cfs_atomic_set(&sbi->ll_agl_total, 0); + sbi->ll_flags |= LL_SBI_AGL_ENABLED; RETURN(sbi); } @@ -162,13 +167,14 @@ static struct dentry_operations ll_d_root_ops = { .d_revalidate = ll_revalidate_nd, }; -static int client_common_fill_super(struct super_block *sb, char *md, char *dt) +static int client_common_fill_super(struct super_block *sb, char *md, char *dt, + struct vfsmount *mnt) { struct inode *root = 0; struct ll_sb_info *sbi = ll_s2sbi(sb); struct obd_device *obd; struct obd_capa *oc = NULL; - struct obd_statfs osfs; + struct obd_statfs *osfs = NULL; struct ptlrpc_request *request = NULL; struct obd_connect_data *data = NULL; struct obd_uuid *uuid; @@ -188,6 +194,12 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) if (data == NULL) RETURN(-ENOMEM); + OBD_ALLOC_PTR(osfs); + if (osfs == NULL) { + OBD_FREE_PTR(data); + RETURN(-ENOMEM); + } + if (proc_lustre_fs_root) { err = lprocfs_register_mountpoint(proc_lustre_fs_root, sb, dt, md); @@ -197,13 +209,13 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) /* indicate the features supported by this client */ data->ocd_connect_flags = OBD_CONNECT_IBITS | OBD_CONNECT_NODEVOH | - OBD_CONNECT_JOIN | OBD_CONNECT_ATTRFID | - OBD_CONNECT_VERSION | OBD_CONNECT_MDS_CAPA | - OBD_CONNECT_OSS_CAPA | OBD_CONNECT_CANCELSET| - OBD_CONNECT_FID | OBD_CONNECT_AT | - OBD_CONNECT_LOV_V3 | OBD_CONNECT_RMT_CLIENT | - OBD_CONNECT_VBR | OBD_CONNECT_FULL20 | - OBD_CONNECT_64BITHASH; + OBD_CONNECT_ATTRFID | + OBD_CONNECT_VERSION | OBD_CONNECT_BRW_SIZE | + OBD_CONNECT_MDS_CAPA | OBD_CONNECT_OSS_CAPA | + OBD_CONNECT_CANCELSET | OBD_CONNECT_FID | + OBD_CONNECT_AT | OBD_CONNECT_LOV_V3 | + OBD_CONNECT_RMT_CLIENT | OBD_CONNECT_VBR | + OBD_CONNECT_FULL20 | OBD_CONNECT_64BITHASH; if (sbi->ll_flags & LL_SBI_SOM_PREVIEW) data->ocd_connect_flags |= OBD_CONNECT_SOM; @@ -243,6 +255,8 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) if (sbi->ll_flags & LL_SBI_RMT_CLIENT) data->ocd_connect_flags |= OBD_CONNECT_RMT_CLIENT_FORCE; + data->ocd_brw_size = PTLRPC_MAX_BRW_SIZE; + err = obd_connect(NULL, &sbi->ll_md_exp, obd, &sbi->ll_sb_uuid, data, NULL); if (err == -EBUSY) { LCONSOLE_ERROR_MSG(0x14f, "An MDT (md %s) is performing " @@ -262,11 +276,35 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) GOTO(out_md, err); } - err = obd_statfs(obd, &osfs, + err = obd_statfs(obd, osfs, cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS), 0); if (err) GOTO(out_md_fid, err); + /* This needs to be after statfs to ensure connect has finished. + * Note that "data" does NOT contain the valid connect reply. + * If connecting to a 1.8 server there will be no LMV device, so + * we can access the MDC export directly and exp_connect_flags will + * be non-zero, but if accessing an upgraded 2.1 server it will + * have the correct flags filled in. + * XXX: fill in the LMV exp_connect_flags from MDC(s). */ + valid = sbi->ll_md_exp->exp_connect_flags & CLIENT_CONNECT_MDT_REQD; + if (sbi->ll_md_exp->exp_connect_flags != 0 && + valid != CLIENT_CONNECT_MDT_REQD) { + char *buf; + + OBD_ALLOC_WAIT(buf, CFS_PAGE_SIZE); + obd_connect_flags2str(buf, CFS_PAGE_SIZE, + valid ^ CLIENT_CONNECT_MDT_REQD, ","); + LCONSOLE_ERROR_MSG(0x170, "Server %s does not support " + "feature(s) needed for correct operation " + "of this client (%s). Please upgrade " + "server or downgrade client.\n", + sbi->ll_md_exp->exp_obd->obd_name, buf); + OBD_FREE(buf, CFS_PAGE_SIZE); + GOTO(out_md, err = -EPROTO); + } + size = sizeof(*data); err = obd_get_info(sbi->ll_md_exp, sizeof(KEY_CONN_DATA), KEY_CONN_DATA, &size, data, NULL); @@ -275,9 +313,9 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) GOTO(out_md, err); } - LASSERT(osfs.os_bsize); - sb->s_blocksize = osfs.os_bsize; - sb->s_blocksize_bits = log2(osfs.os_bsize); + LASSERT(osfs->os_bsize); + sb->s_blocksize = osfs->os_bsize; + sb->s_blocksize_bits = log2(osfs->os_bsize); sb->s_magic = LL_SUPER_MAGIC; /* for bug 11559. in $LINUX/fs/read_write.c, function do_sendfile(): @@ -293,7 +331,7 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) #else sb->s_maxbytes = PAGE_CACHE_MAXBYTES; #endif - sbi->ll_namelen = osfs.os_namelen; + sbi->ll_namelen = osfs->os_namelen; sbi->ll_max_rw_chunk = LL_DEFAULT_MAX_RW_CHUNK; if ((sbi->ll_flags & LL_SBI_USER_XATTR) && @@ -339,6 +377,14 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) sbi->ll_flags |= LL_SBI_OSS_CAPA; } + if (data->ocd_connect_flags & OBD_CONNECT_64BITHASH) + sbi->ll_flags |= LL_SBI_64BIT_HASH; + + if (data->ocd_connect_flags & OBD_CONNECT_BRW_SIZE) + sbi->ll_md_brw_size = data->ocd_brw_size; + else + sbi->ll_md_brw_size = CFS_PAGE_SIZE; + obd = class_name2obd(dt); if (!obd) { CERROR("DT %s: not setup or attached\n", dt); @@ -351,7 +397,8 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) OBD_CONNECT_SRVLOCK | OBD_CONNECT_TRUNCLOCK| OBD_CONNECT_AT | OBD_CONNECT_RMT_CLIENT | OBD_CONNECT_OSS_CAPA | OBD_CONNECT_VBR| - OBD_CONNECT_FULL20; + OBD_CONNECT_FULL20 | OBD_CONNECT_64BITHASH | + OBD_CONNECT_MAXBYTES; if (sbi->ll_flags & LL_SBI_SOM_PREVIEW) data->ocd_connect_flags |= OBD_CONNECT_SOM; @@ -366,8 +413,7 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) if (OBD_FAIL_CHECK(OBD_FAIL_OSC_CKSUM_ADLER_ONLY)) data->ocd_cksum_types = OBD_CKSUM_ADLER; else - /* send the list of supported checksum types */ - data->ocd_cksum_types = OBD_CKSUM_ALL; + data->ocd_cksum_types = cksum_types_supported(); } #ifdef HAVE_LRU_RESIZE_SUPPORT @@ -382,7 +428,8 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) obd->obd_upcall.onu_owner = &sbi->ll_lco; obd->obd_upcall.onu_upcall = cl_ocd_update; - data->ocd_brw_size = PTLRPC_MAX_BRW_PAGES << CFS_PAGE_SHIFT; + + data->ocd_brw_size = PTLRPC_MAX_BRW_SIZE; err = obd_connect(NULL, &sbi->ll_dt_exp, obd, &sbi->ll_sb_uuid, data, NULL); if (err == -EBUSY) { @@ -435,7 +482,7 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) valid |= OBD_MD_FLACL; OBD_ALLOC_PTR(op_data); - if (op_data == NULL) + if (op_data == NULL) GOTO(out_lock_cn_cb, err = -ENOMEM); op_data->op_fid1 = sbi->ll_root_fid; @@ -451,7 +498,6 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) CERROR("md_getattr failed for root: rc = %d\n", err); GOTO(out_lock_cn_cb, err); } - memset(&lmd, 0, sizeof(lmd)); err = md_get_lustre_md(sbi->ll_md_exp, request, sbi->ll_dt_exp, sbi->ll_md_exp, &lmd); if (err) { @@ -500,9 +546,6 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) cl_sb_init(sb); sb->s_root = d_alloc_root(root); - if (data != NULL) - OBD_FREE(data, sizeof(*data)); - sb->s_root->d_op = &ll_d_root_ops; sbi->ll_sdev_orig = sb->s_dev; @@ -515,6 +558,12 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) uuid = obd_get_uuid(sbi->ll_md_exp); if (uuid != NULL) sb->s_dev = get_uuid2int(uuid->uuid, strlen(uuid->uuid)); + sbi->ll_mnt = mnt; + + if (data != NULL) + OBD_FREE_PTR(data); + if (osfs != NULL) + OBD_FREE_PTR(osfs); RETURN(err); out_root: @@ -533,6 +582,8 @@ out_md: out: if (data != NULL) OBD_FREE_PTR(data); + if (osfs != NULL) + OBD_FREE_PTR(osfs); lprocfs_unregister_mountpoint(sbi); return err; } @@ -805,26 +856,54 @@ next: void ll_lli_init(struct ll_inode_info *lli) { lli->lli_inode_magic = LLI_INODE_MAGIC; - cfs_sema_init(&lli->lli_size_sem, 1); - cfs_sema_init(&lli->lli_write_sem, 1); - cfs_init_rwsem(&lli->lli_trunc_sem); lli->lli_flags = 0; - lli->lli_maxbytes = PAGE_CACHE_MAXBYTES; + lli->lli_ioepoch = 0; cfs_spin_lock_init(&lli->lli_lock); - CFS_INIT_LIST_HEAD(&lli->lli_close_list); - lli->lli_inode_magic = LLI_INODE_MAGIC; - cfs_sema_init(&lli->lli_och_sem, 1); - lli->lli_mds_read_och = lli->lli_mds_write_och = NULL; - lli->lli_mds_exec_och = NULL; - lli->lli_open_fd_read_count = lli->lli_open_fd_write_count = 0; - lli->lli_open_fd_exec_count = 0; - CFS_INIT_LIST_HEAD(&lli->lli_dead_list); + lli->lli_posix_acl = NULL; lli->lli_remote_perms = NULL; - lli->lli_rmtperm_utime = 0; cfs_sema_init(&lli->lli_rmtperm_sem, 1); + /* Do not set lli_fid, it has been initialized already. */ + fid_zero(&lli->lli_pfid); + CFS_INIT_LIST_HEAD(&lli->lli_close_list); CFS_INIT_LIST_HEAD(&lli->lli_oss_capas); - cfs_spin_lock_init(&lli->lli_sa_lock); - cfs_sema_init(&lli->lli_readdir_sem, 1); + cfs_atomic_set(&lli->lli_open_count, 0); + lli->lli_mds_capa = NULL; + lli->lli_rmtperm_time = 0; + lli->lli_pending_och = NULL; + lli->lli_mds_read_och = NULL; + lli->lli_mds_write_och = NULL; + lli->lli_mds_exec_och = NULL; + lli->lli_open_fd_read_count = 0; + lli->lli_open_fd_write_count = 0; + lli->lli_open_fd_exec_count = 0; + cfs_sema_init(&lli->lli_och_sem, 1); + cfs_spin_lock_init(&lli->lli_agl_lock); + lli->lli_smd = NULL; + lli->lli_clob = NULL; + + LASSERT(lli->lli_vfs_inode.i_mode != 0); + if (S_ISDIR(lli->lli_vfs_inode.i_mode)) { + cfs_sema_init(&lli->lli_readdir_sem, 1); + lli->lli_opendir_key = NULL; + lli->lli_sai = NULL; + lli->lli_sa_pos = 0; + lli->lli_def_acl = NULL; + cfs_spin_lock_init(&lli->lli_sa_lock); + lli->lli_opendir_pid = 0; + } else { + cfs_sema_init(&lli->lli_size_sem, 1); + lli->lli_size_sem_owner = NULL; + lli->lli_symlink_name = NULL; + lli->lli_maxbytes = PAGE_CACHE_MAXBYTES; + cfs_init_rwsem(&lli->lli_trunc_sem); + cfs_sema_init(&lli->lli_write_sem, 1); + lli->lli_async_rc = 0; + lli->lli_write_rc = 0; + cfs_init_rwsem(&lli->lli_glimpse_sem); + lli->lli_glimpse_time = 0; + CFS_INIT_LIST_HEAD(&lli->lli_agl_list); + lli->lli_agl_index = 0; + } } static inline int ll_bdi_register(struct backing_dev_info *bdi) @@ -832,7 +911,9 @@ static inline int ll_bdi_register(struct backing_dev_info *bdi) #ifdef HAVE_BDI_REGISTER static atomic_t ll_bdi_num = ATOMIC_INIT(0); +#ifdef HAVE_BDI_NAME bdi->name = "lustre"; +#endif return bdi_register(bdi, NULL, "lustre-%d", atomic_inc_return(&ll_bdi_num)); #else @@ -840,26 +921,32 @@ static inline int ll_bdi_register(struct backing_dev_info *bdi) #endif } -int ll_fill_super(struct super_block *sb) +int ll_fill_super(struct super_block *sb, struct vfsmount *mnt) { - struct lustre_profile *lprof; + struct lustre_profile *lprof = NULL; struct lustre_sb_info *lsi = s2lsi(sb); struct ll_sb_info *sbi; char *dt = NULL, *md = NULL; char *profilenm = get_profile_name(sb); - struct config_llog_instance cfg = {0, }; - char ll_instance[sizeof(sb) * 2 + 1]; + struct config_llog_instance *cfg; + /* %p for void* in printf needs 16+2 characters: 0xffffffffffffffff */ + const int instlen = sizeof(cfg->cfg_instance) * 2 + 2; int err; ENTRY; CDEBUG(D_VFSTRACE, "VFS Op: sb %p\n", sb); + OBD_ALLOC_PTR(cfg); + if (cfg == NULL) + RETURN(-ENOMEM); + cfs_module_get(); /* client additional sb info */ lsi->lsi_llsbi = sbi = ll_init_sbi(); if (!sbi) { cfs_module_put(THIS_MODULE); + OBD_FREE_PTR(cfg); RETURN(-ENOMEM); } @@ -883,12 +970,11 @@ int ll_fill_super(struct super_block *sb) /* Generate a string unique to this super, in case some joker tries to mount the same fs at two mount points. Use the address of the super itself.*/ - sprintf(ll_instance, "%p", sb); - cfg.cfg_instance = ll_instance; - cfg.cfg_uuid = lsi->lsi_llsbi->ll_sb_uuid; + cfg->cfg_instance = sb; + cfg->cfg_uuid = lsi->lsi_llsbi->ll_sb_uuid; /* set up client obds */ - err = lustre_process_log(sb, profilenm, &cfg); + err = lustre_process_log(sb, profilenm, cfg); if (err < 0) { CERROR("Unable to process log: %d\n", err); GOTO(out_free, err); @@ -905,31 +991,30 @@ int ll_fill_super(struct super_block *sb) CDEBUG(D_CONFIG, "Found profile %s: mdc=%s osc=%s\n", profilenm, lprof->lp_md, lprof->lp_dt); - OBD_ALLOC(dt, strlen(lprof->lp_dt) + - strlen(ll_instance) + 2); + OBD_ALLOC(dt, strlen(lprof->lp_dt) + instlen + 2); if (!dt) GOTO(out_free, err = -ENOMEM); - sprintf(dt, "%s-%s", lprof->lp_dt, ll_instance); + sprintf(dt, "%s-%p", lprof->lp_dt, cfg->cfg_instance); - OBD_ALLOC(md, strlen(lprof->lp_md) + - strlen(ll_instance) + 2); + OBD_ALLOC(md, strlen(lprof->lp_md) + instlen + 2); if (!md) GOTO(out_free, err = -ENOMEM); - sprintf(md, "%s-%s", lprof->lp_md, ll_instance); + sprintf(md, "%s-%p", lprof->lp_md, cfg->cfg_instance); /* connections, registrations, sb setup */ - err = client_common_fill_super(sb, md, dt); + err = client_common_fill_super(sb, md, dt, mnt); out_free: if (md) - OBD_FREE(md, strlen(md) + 1); + OBD_FREE(md, strlen(lprof->lp_md) + instlen + 2); if (dt) - OBD_FREE(dt, strlen(dt) + 1); + OBD_FREE(dt, strlen(lprof->lp_dt) + instlen + 2); if (err) ll_put_super(sb); else LCONSOLE_WARN("Client %s has started\n", profilenm); + OBD_FREE_PTR(cfg); RETURN(err); } /* ll_fill_super */ @@ -939,22 +1024,19 @@ void lu_context_keys_dump(void); void ll_put_super(struct super_block *sb) { struct config_llog_instance cfg; - char ll_instance[sizeof(sb) * 2 + 1]; struct obd_device *obd; struct lustre_sb_info *lsi = s2lsi(sb); struct ll_sb_info *sbi = ll_s2sbi(sb); char *profilenm = get_profile_name(sb); - int force = 1; - struct obd_device *prev; + int force = 1, next; ENTRY; CDEBUG(D_VFSTRACE, "VFS Op: sb %p - %s\n", sb, profilenm); ll_print_capa_stat(sbi); - sprintf(ll_instance, "%p", sb); - cfg.cfg_instance = ll_instance; - lustre_end_log(sb, NULL, &cfg); + cfg.cfg_instance = sb; + lustre_end_log(sb, profilenm, &cfg); if (sbi->ll_md_exp) { obd = class_exp2obd(sbi->ll_md_exp); @@ -965,9 +1047,9 @@ void ll_put_super(struct super_block *sb) /* We need to set force before the lov_disconnect in lustre_common_put_super, since l_d cleans up osc's as well. */ if (force) { - prev = NULL; + next = 0; while ((obd = class_devices_in_group(&sbi->ll_sb_uuid, - &prev)) != NULL) { + &next)) != NULL) { obd->obd_force = force; } } @@ -977,8 +1059,8 @@ void ll_put_super(struct super_block *sb) client_common_put_super(sb); } - prev = NULL; - while ((obd = class_devices_in_group(&sbi->ll_sb_uuid, &prev)) !=NULL) { + next = 0; + while ((obd = class_devices_in_group(&sbi->ll_sb_uuid, &next)) !=NULL) { class_manual_cleanup(obd); } @@ -997,7 +1079,7 @@ void ll_put_super(struct super_block *sb) cl_env_cache_purge(~0); - LCONSOLE_WARN("client %s umount complete\n", ll_instance); + LCONSOLE_WARN("client %p umount complete\n", cfg.cfg_instance); cfs_module_put(THIS_MODULE); @@ -1049,8 +1131,8 @@ void ll_clear_inode(struct inode *inode) if (S_ISDIR(inode->i_mode)) { /* these should have been cleared in ll_file_release */ - LASSERT(lli->lli_sai == NULL); LASSERT(lli->lli_opendir_key == NULL); + LASSERT(lli->lli_sai == NULL); LASSERT(lli->lli_opendir_pid == 0); } @@ -1069,7 +1151,7 @@ void ll_clear_inode(struct inode *inode) if (lli->lli_mds_read_och) ll_md_real_close(inode, FMODE_READ); - if (lli->lli_symlink_name) { + if (S_ISLNK(inode->i_mode) && lli->lli_symlink_name) { OBD_FREE(lli->lli_symlink_name, strlen(lli->lli_symlink_name) + 1); lli->lli_symlink_name = NULL; @@ -1093,6 +1175,9 @@ void ll_clear_inode(struct inode *inode) lli->lli_inode_magic = LLI_INODE_DEAD; ll_clear_inode_capas(inode); + if (!S_ISDIR(inode->i_mode)) + LASSERT(cfs_list_empty(&lli->lli_agl_list)); + /* * XXX This has to be done before lsm is freed below, because * cl_object still uses inode lsm. @@ -1108,10 +1193,11 @@ void ll_clear_inode(struct inode *inode) EXIT; } -int ll_md_setattr(struct inode *inode, struct md_op_data *op_data, +int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data, struct md_open_data **mod) { struct lustre_md md; + struct inode *inode = dentry->d_inode; struct ll_sb_info *sbi = ll_i2sbi(inode); struct ptlrpc_request *request = NULL; int rc; @@ -1132,7 +1218,7 @@ int ll_md_setattr(struct inode *inode, struct md_op_data *op_data, * Pretend we done everything. */ if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode)) - rc = inode_setattr(inode, &op_data->op_attr); + rc = simple_setattr(dentry, &op_data->op_attr); } else if (rc != -EPERM && rc != -EACCES && rc != -ETXTBSY) { CERROR("md_setattr fails: rc = %d\n", rc); } @@ -1151,7 +1237,7 @@ int ll_md_setattr(struct inode *inode, struct md_op_data *op_data, * above to avoid invoking vmtruncate, otherwise it is important * to call vmtruncate in inode_setattr to update inode->i_size * (bug 6196) */ - rc = inode_setattr(inode, &op_data->op_attr); + rc = simple_setattr(dentry, &op_data->op_attr); /* Extract epoch data if obtained. */ op_data->op_handle = md.body->handle; @@ -1228,8 +1314,9 @@ static int ll_setattr_ost(struct inode *inode, struct iattr *attr) * I don't believe it is possible to get e.g. ATTR_MTIME_SET and ATTR_SIZE * at the same time. */ -int ll_setattr_raw(struct inode *inode, struct iattr *attr) +int ll_setattr_raw(struct dentry *dentry, struct iattr *attr) { + struct inode *inode = dentry->d_inode; struct ll_inode_info *lli = ll_i2info(inode); struct lov_stripe_md *lsm = lli->lli_smd; struct md_op_data *op_data = NULL; @@ -1293,7 +1380,8 @@ int ll_setattr_raw(struct inode *inode, struct iattr *attr) UNLOCK_INODE_MUTEX(inode); if (ia_valid & ATTR_SIZE) UP_WRITE_I_ALLOC_SEM(inode); - cfs_down_write(&lli->lli_trunc_sem); + if (!S_ISDIR(inode->i_mode)) + cfs_down_write(&lli->lli_trunc_sem); LOCK_INODE_MUTEX(inode); if (ia_valid & ATTR_SIZE) DOWN_WRITE_I_ALLOC_SEM(inode); @@ -1305,7 +1393,7 @@ int ll_setattr_raw(struct inode *inode, struct iattr *attr) (ia_valid & (ATTR_SIZE | ATTR_MTIME | ATTR_MTIME_SET))) op_data->op_flags = MF_EPOCH_OPEN; - rc = ll_md_setattr(inode, op_data, &mod); + rc = ll_md_setattr(dentry, op_data, &mod); if (rc) GOTO(out, rc); @@ -1334,7 +1422,8 @@ out: rc1 = ll_setattr_done_writing(inode, op_data, mod); ll_finish_md_op_data(op_data); } - cfs_up_write(&lli->lli_trunc_sem); + if (!S_ISDIR(inode->i_mode)) + cfs_up_write(&lli->lli_trunc_sem); return rc ? rc : rc1; } @@ -1363,7 +1452,7 @@ int ll_setattr(struct dentry *de, struct iattr *attr) !(attr->ia_valid & ATTR_KILL_SGID)) attr->ia_valid |= ATTR_KILL_SGID; - return ll_setattr_raw(de->d_inode, attr); + return ll_setattr_raw(de, attr); } int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs, @@ -1465,6 +1554,8 @@ void ll_inode_size_lock(struct inode *inode, int lock_lsm) struct ll_inode_info *lli; struct lov_stripe_md *lsm; + LASSERT(!S_ISDIR(inode->i_mode)); + lli = ll_i2info(inode); LASSERT(lli->lli_size_sem_owner != current); cfs_down(&lli->lli_size_sem); @@ -1502,6 +1593,8 @@ void ll_update_inode(struct inode *inode, struct lustre_md *md) LASSERT ((lsm != NULL) == ((body->valid & OBD_MD_FLEASIZE) != 0)); if (lsm != NULL) { + LASSERT(S_ISREG(inode->i_mode)); + cfs_down(&lli->lli_och_sem); if (lli->lli_smd == NULL) { if (lsm->lsm_magic != LOV_MAGIC_V1 && @@ -1511,12 +1604,14 @@ void ll_update_inode(struct inode *inode, struct lustre_md *md) } CDEBUG(D_INODE, "adding lsm %p to inode %lu/%u(%p)\n", lsm, inode->i_ino, inode->i_generation, inode); - /* cl_inode_init must go before lli_smd or a race is - * possible where client thinks the file has stripes, + /* cl_file_inode_init must go before lli_smd or a race + * is possible where client thinks the file has stripes, * but lov raid0 is not setup yet and parallel e.g. * glimpse would try to use uninitialized lov */ - cl_inode_init(inode, md); + cl_file_inode_init(inode, md); + cfs_spin_lock(&lli->lli_lock); lli->lli_smd = lsm; + cfs_spin_unlock(&lli->lli_lock); cfs_up(&lli->lli_och_sem); lli->lli_maxbytes = lsm->lsm_maxbytes; if (lli->lli_maxbytes > PAGE_CACHE_MAXBYTES) @@ -1629,14 +1724,14 @@ void ll_update_inode(struct inode *inode, struct lustre_md *md) if (lli->lli_flags & (LLIF_DONE_WRITING | LLIF_EPOCH_PENDING | LLIF_SOM_DIRTY)) { - CERROR("ino %lu flags %lu still has " + CERROR("ino %lu flags %u still has " "size authority! do not trust " "the size got from MDS\n", inode->i_ino, lli->lli_flags); } else { /* Use old size assignment to avoid * deadlock bz14138 & bz14326 */ - inode->i_size = body->size; + i_size_write(inode, body->size); lli->lli_flags |= LLIF_MDS_SIZE_LOCK; } ldlm_lock_decref(&lockh, mode); @@ -1644,7 +1739,7 @@ void ll_update_inode(struct inode *inode, struct lustre_md *md) } else { /* Use old size assignment to avoid * deadlock bz14138 & bz14326 */ - inode->i_size = body->size; + i_size_write(inode, body->size); CDEBUG(D_VFSTRACE, "inode=%lu, updating i_size %llu\n", inode->i_ino, (unsigned long long)body->size); @@ -1670,10 +1765,8 @@ void ll_read_inode2(struct inode *inode, void *opaque) struct ll_inode_info *lli = ll_i2info(inode); ENTRY; - CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p)\n", - inode->i_ino, inode->i_generation, inode); - - ll_lli_init(lli); + CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p)\n", + PFID(&lli->lli_fid), inode); LASSERT(!lli->lli_smd); @@ -1724,11 +1817,29 @@ void ll_delete_inode(struct inode *inode) ENTRY; rc = obd_fid_delete(sbi->ll_md_exp, ll_inode2fid(inode)); - if (rc) { + if (rc) CERROR("fid_delete() failed, rc %d\n", rc); - } + truncate_inode_pages(&inode->i_data, 0); + + /* Workaround for LU-118 */ + if (inode->i_data.nrpages) { + TREE_READ_LOCK_IRQ(&inode->i_data); + TREE_READ_UNLOCK_IRQ(&inode->i_data); + LASSERTF(inode->i_data.nrpages == 0, + "inode=%lu/%u(%p) nrpages=%lu, see " + "http://jira.whamcloud.com/browse/LU-118\n", + inode->i_ino, inode->i_generation, inode, + inode->i_data.nrpages); + } + /* Workaround end */ + +#ifdef HAVE_SBOPS_EVICT_INODE + ll_clear_inode(inode); + end_writeback(inode); +#else clear_inode(inode); +#endif EXIT; } @@ -1749,8 +1860,8 @@ int ll_iocontrol(struct inode *inode, struct file *file, op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL, 0, 0, LUSTRE_OPC_ANY, NULL); - if (op_data == NULL) - RETURN(-ENOMEM); + if (IS_ERR(op_data)) + RETURN(PTR_ERR(op_data)); op_data->op_valid = OBD_MD_FLFLAGS; rc = md_getattr(sbi->ll_md_exp, op_data, &req); @@ -1960,8 +2071,6 @@ int ll_prep_inode(struct inode **inode, LASSERT(*inode || sb); sbi = sb ? ll_s2sbi(sb) : ll_i2sbi(*inode); - memset(&md, 0, sizeof(struct lustre_md)); - rc = md_get_lustre_md(sbi->ll_md_exp, req, sbi->ll_dt_exp, sbi->ll_md_exp, &md); if (rc) @@ -2028,7 +2137,7 @@ int ll_obd_statfs(struct inode *inode, void *arg) GOTO(out_statfs, rc = -EINVAL); memcpy(&type, data->ioc_inlbuf1, sizeof(__u32)); - if (type == LL_STATFS_MDC) + if (type == LL_STATFS_LMV) exp = sbi->ll_md_exp; else if (type == LL_STATFS_LOV) exp = sbi->ll_dt_exp; @@ -2114,6 +2223,21 @@ struct md_op_data * ll_prep_md_op_data(struct md_op_data *op_data, op_data->op_mds = 0; op_data->op_data = data; + /* If the file is being opened after mknod() (normally due to NFS) + * try to use the default stripe data from parent directory for + * allocating OST objects. Try to pass the parent FID to MDS. */ + if (opc == LUSTRE_OPC_CREATE && i1 == i2 && S_ISREG(i2->i_mode) && + ll_i2info(i2)->lli_smd == NULL) { + struct ll_inode_info *lli = ll_i2info(i2); + + cfs_spin_lock(&lli->lli_lock); + if (likely(lli->lli_smd == NULL && + !fid_is_zero(&lli->lli_pfid))) + op_data->op_fid1 = lli->lli_pfid; + cfs_spin_unlock(&lli->lli_lock); + /** We ignore parent's capability temporary. */ + } + return op_data; } @@ -2148,3 +2272,29 @@ int ll_show_options(struct seq_file *seq, struct vfsmount *vfs) RETURN(0); } + +/** + * Get obd name by cmd, and copy out to user space + */ +int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg) +{ + struct ll_sb_info *sbi = ll_i2sbi(inode); + struct obd_device *obd; + ENTRY; + + if (cmd == OBD_IOC_GETDTNAME) + obd = class_exp2obd(sbi->ll_dt_exp); + else if (cmd == OBD_IOC_GETMDNAME) + obd = class_exp2obd(sbi->ll_md_exp); + else + RETURN(-EINVAL); + + if (!obd) + RETURN(-ENOENT); + + if (cfs_copy_to_user((void *)arg, obd->obd_name, + strlen(obd->obd_name) + 1)) + RETURN(-EFAULT); + + RETURN(0); +}