X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fllite%2Fllite_lib.c;h=f97250b41a645353eb035518c49952bc53b00bcb;hp=66a57eeafe9b42348c1f0d6e247141ae20d84527;hb=ed2e4d8205e5b6bc9dc2ad8319ad7666e49e5dfe;hpb=0993ba6915f7f6529526eb1672a345a10b86a02d diff --git a/lustre/llite/llite_lib.c b/lustre/llite/llite_lib.c index 66a57ee..f97250b 100644 --- a/lustre/llite/llite_lib.c +++ b/lustre/llite/llite_lib.c @@ -26,10 +26,13 @@ * GPL HEADER END */ /* - * Copyright 2008 Sun Microsystems, Inc. All rights reserved + * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved. * 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. * @@ -42,7 +45,6 @@ #include #include -#include #include #include @@ -62,8 +64,13 @@ cfs_mem_cache_t *ll_file_data_slab; CFS_LIST_HEAD(ll_super_blocks); cfs_spinlock_t ll_sb_lock = CFS_SPIN_LOCK_UNLOCKED; +#ifndef MS_HAS_NEW_AOPS extern struct address_space_operations ll_aops; extern struct address_space_operations ll_dir_aops; +#else +extern struct address_space_operations_ext ll_aops; +extern struct address_space_operations_ext ll_dir_aops; +#endif #ifndef log2 #define log2(n) cfs_ffz(~(n)) @@ -116,7 +123,7 @@ static struct ll_sb_info *ll_init_sbi(void) cfs_list_add_tail(&sbi->ll_list, &ll_super_blocks); cfs_spin_unlock(&ll_sb_lock); -#ifdef ENABLE_LLITE_CHECKSUM +#ifdef ENABLE_CHECKSUM sbi->ll_flags |= LL_SBI_CHECKSUM; #endif @@ -158,13 +165,14 @@ 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; @@ -184,6 +192,12 @@ 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); @@ -193,12 +207,13 @@ 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_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; @@ -222,6 +237,9 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) /* force vfs to use lustre handler for flock() calls - bug 10743 */ sb->s_flags |= MS_FLOCK_LOCK; #endif +#ifdef MS_HAS_NEW_AOPS + sb->s_flags |= MS_HAS_NEW_AOPS; +#endif if (sbi->ll_flags & LL_SBI_FLOCK) sbi->ll_fop = &ll_file_operations_flock; @@ -235,6 +253,8 @@ 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 " @@ -254,10 +274,35 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) GOTO(out_md, err); } - err = obd_statfs(obd, &osfs, cfs_time_current_64() - CFS_HZ, 0); + 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); @@ -266,9 +311,9 @@ 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(): @@ -284,7 +329,7 @@ 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) && @@ -330,6 +375,14 @@ 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); @@ -342,7 +395,8 @@ 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; @@ -373,7 +427,8 @@ 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) { @@ -413,7 +468,9 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) CDEBUG(D_SUPER, "rootfid "DFID"\n", PFID(&sbi->ll_root_fid)); sb->s_op = &lustre_super_operations; +#if THREAD_SIZE >= 8192 sb->s_export_op = &lustre_export_operations; +#endif /* make root inode * XXX: move this to after cbd setup? */ @@ -424,7 +481,7 @@ 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; @@ -450,7 +507,7 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) } LASSERT(fid_is_sane(&sbi->ll_root_fid)); - root = ll_iget(sb, cl_fid_build_ino(&sbi->ll_root_fid), &lmd); + root = ll_iget(sb, cl_fid_build_ino(&sbi->ll_root_fid, 0), &lmd); md_free_lustre_md(sbi->ll_md_exp, &lmd); ptlrpc_req_finished(request); @@ -489,9 +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; @@ -504,6 +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: @@ -522,6 +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; } @@ -711,17 +773,25 @@ static int ll_options(char *options, int *flags) *flags &= ~tmp; goto next; } +#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2,5,50,0) tmp = ll_set_opt("acl", s1, LL_SBI_ACL); if (tmp) { /* Ignore deprecated mount option. The client will * always try to mount with ACL support, whether this * is used depends on whether server supports it. */ + LCONSOLE_ERROR_MSG(0x152, "Ignoring deprecated " + "mount option 'acl'.\n"); goto next; } tmp = ll_set_opt("noacl", s1, LL_SBI_ACL); if (tmp) { + LCONSOLE_ERROR_MSG(0x152, "Ignoring deprecated " + "mount option 'noacl'.\n"); goto next; } +#else +#warning "{no}acl options have been deprecated since 1.8, please remove them" +#endif tmp = ll_set_opt("remote_client", s1, LL_SBI_RMT_CLIENT); if (tmp) { *flags |= tmp; @@ -763,6 +833,11 @@ static int ll_options(char *options, int *flags) *flags |= tmp; goto next; } + tmp = ll_set_opt("32bitapi", s1, LL_SBI_32BIT_API); + if (tmp) { + *flags |= tmp; + goto next; + } LCONSOLE_ERROR_MSG(0x152, "Unknown option '%s', won't mount.\n", s1); @@ -783,7 +858,7 @@ 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_sema_init(&lli->lli_trunc_sem, 1); + cfs_init_rwsem(&lli->lli_trunc_sem); lli->lli_flags = 0; lli->lli_maxbytes = PAGE_CACHE_MAXBYTES; cfs_spin_lock_init(&lli->lli_lock); @@ -800,29 +875,49 @@ void ll_lli_init(struct ll_inode_info *lli) cfs_sema_init(&lli->lli_rmtperm_sem, 1); CFS_INIT_LIST_HEAD(&lli->lli_oss_capas); cfs_spin_lock_init(&lli->lli_sa_lock); - CFS_INIT_LIST_HEAD(&lli->lli_sa_dentry); + cfs_sema_init(&lli->lli_readdir_sem, 1); + fid_zero(&lli->lli_pfid); +} + +static inline int ll_bdi_register(struct backing_dev_info *bdi) +{ +#ifdef HAVE_BDI_REGISTER + static atomic_t ll_bdi_num = ATOMIC_INIT(0); + + bdi->name = "lustre"; + return bdi_register(bdi, NULL, "lustre-%d", + atomic_inc_return(&ll_bdi_num)); +#else + return 0; +#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); } @@ -830,20 +925,33 @@ int ll_fill_super(struct super_block *sb) if (err) GOTO(out_free, err); + err = ll_bdi_init(&lsi->lsi_bdi); + if (err) + GOTO(out_free, err); + lsi->lsi_flags |= LSI_BDI_INITIALIZED; + lsi->lsi_bdi.capabilities = BDI_CAP_MAP_COPY; + err = ll_bdi_register(&lsi->lsi_bdi); + if (err) + GOTO(out_free, err); + +#ifdef HAVE_SB_BDI + sb->s_bdi = &lsi->lsi_bdi; +#endif + /* 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); } + /* Profile set with LCFG_MOUNTOPT so we can find our mdc and osc obds */ lprof = class_get_profile(profilenm); if (lprof == NULL) { LCONSOLE_ERROR_MSG(0x156, "The client profile '%s' could not be" @@ -854,31 +962,30 @@ 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 */ @@ -888,7 +995,6 @@ 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); @@ -900,9 +1006,8 @@ void ll_put_super(struct super_block *sb) 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); @@ -933,6 +1038,11 @@ void ll_put_super(struct super_block *sb) if (profilenm) class_del_profile(profilenm); + if (lsi->lsi_flags & LSI_BDI_INITIALIZED) { + ll_bdi_destroy(&lsi->lsi_bdi); + lsi->lsi_flags &= ~LSI_BDI_INITIALIZED; + } + ll_free_sbi(sb); lsi->lsi_llsbi = NULL; @@ -940,7 +1050,7 @@ 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); @@ -958,11 +1068,10 @@ struct inode *ll_inode_from_lock(struct ldlm_lock *lock) inode = igrab(lock->l_ast_data); } else { inode = lock->l_ast_data; - ldlm_lock_debug(NULL, inode->i_state & I_FREEING ? - D_INFO : D_WARNING, - lock, __FILE__, __func__, __LINE__, - "l_ast_data %p is bogus: magic %08x", - lock->l_ast_data, lli->lli_inode_magic); + LDLM_DEBUG_LIMIT(inode->i_state & I_FREEING ? D_INFO : + D_WARNING, lock, "l_ast_data %p is " + "bogus: magic %08x", lock->l_ast_data, + lli->lli_inode_magic); inode = NULL; } } @@ -1237,7 +1346,7 @@ 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(&lli->lli_trunc_sem); + cfs_down_write(&lli->lli_trunc_sem); LOCK_INODE_MUTEX(inode); if (ia_valid & ATTR_SIZE) DOWN_WRITE_I_ALLOC_SEM(inode); @@ -1261,13 +1370,15 @@ int ll_setattr_raw(struct inode *inode, struct iattr *attr) if (ia_valid & ATTR_SIZE) attr->ia_valid |= ATTR_SIZE; - if ((ia_valid & ATTR_SIZE) | - ((ia_valid | ATTR_ATIME | ATTR_ATIME_SET) && - LTIME_S(attr->ia_atime) < LTIME_S(attr->ia_ctime)) || - ((ia_valid | ATTR_MTIME | ATTR_MTIME_SET) && - LTIME_S(attr->ia_mtime) < LTIME_S(attr->ia_ctime))) - /* perform truncate and setting mtime/atime to past under PW - * 0:EOF extent lock (new_size:EOF for truncate) */ + if (ia_valid & (ATTR_SIZE | + ATTR_ATIME | ATTR_ATIME_SET | + ATTR_MTIME | ATTR_MTIME_SET)) + /* on truncate and utimes send attributes to osts, setting + * mtime/atime to past will be performed under PW 0:EOF extent + * lock (new_size:EOF for truncate) + * it may seem excessive to send mtime/atime updates to osts + * when not setting times to past, but it is necessary due to + * possible time de-synchronization */ rc = ll_setattr_ost(inode, attr); EXIT; out: @@ -1276,22 +1387,31 @@ out: rc1 = ll_setattr_done_writing(inode, op_data, mod); ll_finish_md_op_data(op_data); } - cfs_up(&lli->lli_trunc_sem); + cfs_up_write(&lli->lli_trunc_sem); return rc ? rc : rc1; } int ll_setattr(struct dentry *de, struct iattr *attr) { + int mode = de->d_inode->i_mode; + if ((attr->ia_valid & (ATTR_CTIME|ATTR_SIZE|ATTR_MODE)) == - (ATTR_CTIME|ATTR_SIZE|ATTR_MODE)) + (ATTR_CTIME|ATTR_SIZE|ATTR_MODE)) attr->ia_valid |= MDS_OPEN_OWNEROVERRIDE; - if ((de->d_inode->i_mode & S_ISUID) && + if (((attr->ia_valid & (ATTR_MODE|ATTR_FORCE|ATTR_SIZE)) == + (ATTR_SIZE|ATTR_MODE)) && + (((mode & S_ISUID) && !(attr->ia_mode & S_ISUID)) || + (((mode & (S_ISGID|S_IXGRP)) == (S_ISGID|S_IXGRP)) && + !(attr->ia_mode & S_ISGID)))) + attr->ia_valid |= ATTR_FORCE; + + if ((mode & S_ISUID) && !(attr->ia_mode & S_ISUID) && !(attr->ia_valid & ATTR_KILL_SUID)) attr->ia_valid |= ATTR_KILL_SUID; - if (((de->d_inode->i_mode & (S_ISGID|S_IXGRP)) == (S_ISGID|S_IXGRP)) && + if (((mode & (S_ISGID|S_IXGRP)) == (S_ISGID|S_IXGRP)) && !(attr->ia_mode & S_ISGID) && !(attr->ia_valid & ATTR_KILL_SGID)) attr->ia_valid |= ATTR_KILL_SGID; @@ -1363,10 +1483,10 @@ int ll_statfs(struct dentry *de, struct kstatfs *sfs) CDEBUG(D_VFSTRACE, "VFS Op: at "LPU64" jiffies\n", get_jiffies_64()); ll_stats_ops_tally(ll_s2sbi(sb), LPROC_LL_STAFS, 1); - /* For now we will always get up-to-date statfs values, but in the - * future we may allow some amount of caching on the client (e.g. - * from QOS or lprocfs updates). */ - rc = ll_statfs_internal(sb, &osfs, cfs_time_current_64() - 1, 0); + /* Some amount of caching on the client is allowed */ + rc = ll_statfs_internal(sb, &osfs, + cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS), + 0); if (rc) return rc; @@ -1449,7 +1569,9 @@ void ll_update_inode(struct inode *inode, struct lustre_md *md) * but lov raid0 is not setup yet and parallel e.g. * glimpse would try to use uninitialized lov */ cl_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) @@ -1486,7 +1608,7 @@ void ll_update_inode(struct inode *inode, struct lustre_md *md) cfs_spin_unlock(&lli->lli_lock); } #endif - inode->i_ino = cl_fid_build_ino(&body->fid1); + inode->i_ino = cl_fid_build_ino(&body->fid1, 0); inode->i_generation = cl_fid_build_gen(&body->fid1); if (body->valid & OBD_MD_FLATIME) { @@ -1578,6 +1700,9 @@ void ll_update_inode(struct inode *inode, struct lustre_md *md) /* Use old size assignment to avoid * deadlock bz14138 & bz14326 */ inode->i_size = body->size; + + CDEBUG(D_VFSTRACE, "inode=%lu, updating i_size %llu\n", + inode->i_ino, (unsigned long long)body->size); } if (body->valid & OBD_MD_FLBLOCKS) @@ -1594,15 +1719,6 @@ void ll_update_inode(struct inode *inode, struct lustre_md *md) } } -static struct backing_dev_info ll_backing_dev_info = { - .ra_pages = 0, /* No readahead */ -#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,12)) - .capabilities = 0, /* Does contribute to dirty memory */ -#else - .memory_backed = 0, /* Does contribute to dirty memory */ -#endif -}; - void ll_read_inode2(struct inode *inode, void *opaque) { struct lustre_md *md = opaque; @@ -1628,16 +1744,20 @@ void ll_read_inode2(struct inode *inode, void *opaque) /* OIDEBUG(inode); */ + /* initializing backing dev info. */ + inode->i_mapping->backing_dev_info = &s2lsi(inode->i_sb)->lsi_bdi; + + if (S_ISREG(inode->i_mode)) { struct ll_sb_info *sbi = ll_i2sbi(inode); inode->i_op = &ll_file_inode_operations; inode->i_fop = sbi->ll_fop; - inode->i_mapping->a_ops = &ll_aops; + inode->i_mapping->a_ops = (struct address_space_operations *)&ll_aops; EXIT; } else if (S_ISDIR(inode->i_mode)) { inode->i_op = &ll_dir_inode_operations; inode->i_fop = &ll_dir_operations; - inode->i_mapping->a_ops = &ll_dir_aops; + inode->i_mapping->a_ops = (struct address_space_operations *)&ll_dir_aops; EXIT; } else if (S_ISLNK(inode->i_mode)) { inode->i_op = &ll_fast_symlink_inode_operations; @@ -1648,9 +1768,6 @@ void ll_read_inode2(struct inode *inode, void *opaque) init_special_inode(inode, inode->i_mode, kdev_t_to_nr(inode->i_rdev)); - /* initializing backing dev info. */ - inode->i_mapping->backing_dev_info = &ll_backing_dev_info; - EXIT; } } @@ -1662,10 +1779,23 @@ 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 */ + clear_inode(inode); EXIT; @@ -1741,14 +1871,13 @@ int ll_iocontrol(struct inode *inode, struct file *file, } oinfo.oi_oa->o_id = lsm->lsm_object_id; - oinfo.oi_oa->o_gr = lsm->lsm_object_gr; + oinfo.oi_oa->o_seq = lsm->lsm_object_seq; oinfo.oi_oa->o_flags = flags; oinfo.oi_oa->o_valid = OBD_MD_FLID | OBD_MD_FLFLAGS | OBD_MD_FLGROUP; oinfo.oi_capa = ll_mdscapa_get(inode); - obdo_from_inode(oinfo.oi_oa, inode, - OBD_MD_FLFID | OBD_MD_FLGENER); + &ll_i2info(inode)->lli_fid, 0); rc = obd_setattr_rqset(sbi->ll_dt_exp, &oinfo, NULL); capa_put(oinfo.oi_capa); OBDO_FREE(oinfo.oi_oa); @@ -1917,7 +2046,7 @@ int ll_prep_inode(struct inode **inode, */ LASSERT(fid_is_sane(&md.body->fid1)); - *inode = ll_iget(sb, cl_fid_build_ino(&md.body->fid1), &md); + *inode = ll_iget(sb, cl_fid_build_ino(&md.body->fid1, 0), &md); if (*inode == NULL || IS_ERR(*inode)) { if (md.lsm) obd_free_memmd(sbi->ll_dt_exp, &md.lsm); @@ -2053,6 +2182,21 @@ 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; } @@ -2082,9 +2226,6 @@ int ll_show_options(struct seq_file *seq, struct vfsmount *vfs) if (sbi->ll_flags & LL_SBI_USER_XATTR) seq_puts(seq, ",user_xattr"); - if (sbi->ll_flags & LL_SBI_ACL) - seq_puts(seq, ",acl"); - if (sbi->ll_flags & LL_SBI_LAZYSTATFS) seq_puts(seq, ",lazystatfs");