X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fllite%2Fllite_lib.c;h=c8bba124584ab8afee7bbb498cdb93053f91dffc;hp=76bc5d0dfd19cc61451fe73979e19a7a4f3dffa9;hb=9496591a7f31e8766ee1124725fbab3841d8baef;hpb=f957caaa9b2afa4777bf7dbb573354421811681a diff --git a/lustre/llite/llite_lib.c b/lustre/llite/llite_lib.c index 76bc5d0..c8bba12 100644 --- a/lustre/llite/llite_lib.c +++ b/lustre/llite/llite_lib.c @@ -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. @@ -61,14 +59,12 @@ cfs_mem_cache_t *ll_file_data_slab; CFS_LIST_HEAD(ll_super_blocks); -cfs_spinlock_t ll_sb_lock = CFS_SPIN_LOCK_UNLOCKED; +DEFINE_SPINLOCK(ll_sb_lock); #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 @@ -77,8 +73,9 @@ extern struct address_space_operations_ext ll_dir_aops; static struct ll_sb_info *ll_init_sbi(void) { - struct ll_sb_info *sbi = NULL; - unsigned long pages; + struct ll_sb_info *sbi = NULL; + unsigned long pages; + unsigned long lru_page_max; struct sysinfo si; class_uuid_t uuid; int i; @@ -98,13 +95,20 @@ static struct ll_sb_info *ll_init_sbi(void) pages = si.totalram - si.totalhigh; if (pages >> (20 - CFS_PAGE_SHIFT) < 512) { #ifdef HAVE_BGL_SUPPORT - sbi->ll_async_page_max = pages / 4; + lru_page_max = pages / 4; #else - sbi->ll_async_page_max = pages / 2; + lru_page_max = pages / 2; #endif - } else { - sbi->ll_async_page_max = (pages / 4) * 3; - } + } else { + lru_page_max = (pages / 4) * 3; + } + + /* initialize lru data */ + cfs_atomic_set(&sbi->ll_cache.ccc_users, 0); + sbi->ll_cache.ccc_lru_max = lru_page_max; + cfs_atomic_set(&sbi->ll_cache.ccc_lru_left, lru_page_max); + cfs_spin_lock_init(&sbi->ll_cache.ccc_lru_lock); + CFS_INIT_LIST_HEAD(&sbi->ll_cache.ccc_lru); sbi->ll_ra_info.ra_max_pages_per_file = min(pages / 32, SBI_DEFAULT_READAHEAD_MAX); @@ -122,6 +126,7 @@ 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); + sbi->ll_flags |= LL_SBI_VERBOSE; #ifdef ENABLE_CHECKSUM sbi->ll_flags |= LL_SBI_CHECKSUM; #endif @@ -214,7 +219,9 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt, 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_FULL20 | OBD_CONNECT_64BITHASH| + OBD_CONNECT_EINPROGRESS | + OBD_CONNECT_JOBSTATS; if (sbi->ll_flags & LL_SBI_SOM_PREVIEW) data->ocd_connect_flags |= OBD_CONNECT_SOM; @@ -226,6 +233,12 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt, #ifdef CONFIG_FS_POSIX_ACL data->ocd_connect_flags |= OBD_CONNECT_ACL; #endif + + if (OBD_FAIL_CHECK(OBD_FAIL_MDC_LIGHTWEIGHT)) + /* flag mdc connection as lightweight, only used for test + * purpose, use with care */ + data->ocd_connect_flags |= OBD_CONNECT_LIGHTWEIGHT; + data->ocd_ibits_known = MDS_INODELOCK_FULL; data->ocd_version = LUSTRE_VERSION_CODE; @@ -275,7 +288,7 @@ 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(NULL, sbi->ll_md_exp, osfs, cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS), 0); if (err) GOTO(out_md_fid, err); @@ -305,7 +318,7 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt, } size = sizeof(*data); - err = obd_get_info(sbi->ll_md_exp, sizeof(KEY_CONN_DATA), + err = obd_get_info(NULL, sbi->ll_md_exp, sizeof(KEY_CONN_DATA), KEY_CONN_DATA, &size, data, NULL); if (err) { CERROR("Get connect data failed: %d \n", err); @@ -371,6 +384,11 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt, else sbi->ll_md_brw_size = CFS_PAGE_SIZE; + if (data->ocd_connect_flags & OBD_CONNECT_LAYOUTLOCK) { + LCONSOLE_INFO("Layout lock feature supported.\n"); + sbi->ll_flags |= LL_SBI_LAYOUT_LOCK; + } + obd = class_name2obd(dt); if (!obd) { CERROR("DT %s: not setup or attached\n", dt); @@ -384,7 +402,9 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt, OBD_CONNECT_AT | OBD_CONNECT_RMT_CLIENT | OBD_CONNECT_OSS_CAPA | OBD_CONNECT_VBR| OBD_CONNECT_FULL20 | OBD_CONNECT_64BITHASH | - OBD_CONNECT_MAXBYTES; + OBD_CONNECT_MAXBYTES | + OBD_CONNECT_EINPROGRESS | + OBD_CONNECT_JOBSTATS; if (sbi->ll_flags & LL_SBI_SOM_PREVIEW) data->ocd_connect_flags |= OBD_CONNECT_SOM; @@ -396,10 +416,10 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt, * agreement on the supported algorithms at connect time */ data->ocd_connect_flags |= OBD_CONNECT_CKSUM; - if (OBD_FAIL_CHECK(OBD_FAIL_OSC_CKSUM_ADLER_ONLY)) - data->ocd_cksum_types = OBD_CKSUM_ADLER; - else - data->ocd_cksum_types = cksum_types_supported(); + if (OBD_FAIL_CHECK(OBD_FAIL_OSC_CKSUM_ADLER_ONLY)) + data->ocd_cksum_types = OBD_CKSUM_ADLER; + else + data->ocd_cksum_types = cksum_types_supported_client(); } #ifdef HAVE_LRU_RESIZE_SUPPORT @@ -455,7 +475,7 @@ 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 +#if THREAD_SIZE >= 8192 /*b=17630*/ sb->s_export_op = &lustre_export_operations; #endif @@ -526,13 +546,23 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt, #endif checksum = sbi->ll_flags & LL_SBI_CHECKSUM; - err = obd_set_info_async(sbi->ll_dt_exp, sizeof(KEY_CHECKSUM), + err = obd_set_info_async(NULL, sbi->ll_dt_exp, sizeof(KEY_CHECKSUM), KEY_CHECKSUM, sizeof(checksum), &checksum, NULL); cl_sb_init(sb); - sb->s_root = d_alloc_root(root); - sb->s_root->d_op = &ll_d_root_ops; + err = obd_set_info_async(NULL, sbi->ll_dt_exp, sizeof(KEY_CACHE_SET), + KEY_CACHE_SET, sizeof(sbi->ll_cache), + &sbi->ll_cache, NULL); + + sb->s_root = d_alloc_root(root); +#ifdef HAVE_DCACHE_LOCK + sb->s_root->d_op = &ll_d_root_ops; +#else + /* kernel >= 2.6.38 store dentry operations in sb->s_d_op. */ + d_set_d_op(sb->s_root, &ll_d_root_ops); + sb->s_d_op = &ll_d_ops; +#endif sbi->ll_sdev_orig = sb->s_dev; @@ -544,7 +574,6 @@ 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); @@ -580,7 +609,7 @@ int ll_get_max_mdsize(struct ll_sb_info *sbi, int *lmmsize) *lmmsize = obd_size_diskmd(sbi->ll_dt_exp, NULL); size = sizeof(int); - rc = obd_get_info(sbi->ll_md_exp, sizeof(KEY_MAX_EASIZE), + rc = obd_get_info(NULL, sbi->ll_md_exp, sizeof(KEY_MAX_EASIZE), KEY_MAX_EASIZE, &size, lmmsize, NULL); if (rc) CERROR("Get max mdsize error rc %d \n", rc); @@ -617,7 +646,7 @@ void lustre_dump_dentry(struct dentry *dentry, int recur) " flags=0x%x, fsdata=%p, %d subdirs\n", dentry, dentry->d_name.len, dentry->d_name.name, dentry->d_parent->d_name.len, dentry->d_parent->d_name.name, - dentry->d_parent, dentry->d_inode, atomic_read(&dentry->d_count), + dentry->d_parent, dentry->d_inode, d_refcount(dentry), dentry->d_flags, dentry->d_fsdata, subdirs); if (dentry->d_inode != NULL) ll_dump_inode(dentry->d_inode); @@ -625,10 +654,10 @@ void lustre_dump_dentry(struct dentry *dentry, int recur) if (recur == 0) return; - list_for_each(tmp, &dentry->d_subdirs) { - struct dentry *d = list_entry(tmp, struct dentry, d_child); - lustre_dump_dentry(d, recur - 1); - } + list_for_each(tmp, &dentry->d_subdirs) { + struct dentry *d = list_entry(tmp, struct dentry, d_u.d_child); + lustre_dump_dentry(d, recur - 1); + } } void client_common_put_super(struct super_block *sb) @@ -759,7 +788,7 @@ static int ll_options(char *options, int *flags) *flags &= ~tmp; goto next; } -#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2,5,50,0) +#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 @@ -783,6 +812,16 @@ static int ll_options(char *options, int *flags) *flags |= tmp; goto next; } + tmp = ll_set_opt("user_fid2path", s1, LL_SBI_USER_FID2PATH); + if (tmp) { + *flags |= tmp; + goto next; + } + tmp = ll_set_opt("nouser_fid2path", s1, LL_SBI_USER_FID2PATH); + if (tmp) { + *flags &= ~tmp; + goto next; + } tmp = ll_set_opt("checksum", s1, LL_SBI_CHECKSUM); if (tmp) { @@ -824,7 +863,16 @@ static int ll_options(char *options, int *flags) *flags |= tmp; goto next; } - + tmp = ll_set_opt("verbose", s1, LL_SBI_VERBOSE); + if (tmp) { + *flags |= tmp; + goto next; + } + tmp = ll_set_opt("noverbose", s1, LL_SBI_VERBOSE); + if (tmp) { + *flags &= ~tmp; + goto next; + } LCONSOLE_ERROR_MSG(0x152, "Unknown option '%s', won't mount.\n", s1); RETURN(-EINVAL); @@ -865,7 +913,7 @@ void ll_lli_init(struct ll_inode_info *lli) lli->lli_open_fd_exec_count = 0; cfs_mutex_init(&lli->lli_och_mutex); cfs_spin_lock_init(&lli->lli_agl_lock); - lli->lli_smd = NULL; + lli->lli_has_smd = false; lli->lli_clob = NULL; LASSERT(lli->lli_vfs_inode.i_mode != 0); @@ -873,7 +921,6 @@ void ll_lli_init(struct ll_inode_info *lli) cfs_mutex_init(&lli->lli_readdir_mutex); 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; @@ -883,13 +930,13 @@ void ll_lli_init(struct ll_inode_info *lli) lli->lli_symlink_name = NULL; cfs_init_rwsem(&lli->lli_trunc_sem); cfs_mutex_init(&lli->lli_write_mutex); - 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; - } + 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; + lli->lli_async_rc = 0; + } + cfs_mutex_init(&lli->lli_layout_mutex); } static inline int ll_bdi_register(struct backing_dev_info *bdi) @@ -958,7 +1005,7 @@ int ll_fill_super(struct super_block *sb, struct vfsmount *mnt) Use the address of the super itself.*/ cfg->cfg_instance = sb; cfg->cfg_uuid = lsi->lsi_llsbi->ll_sb_uuid; - + cfg->cfg_callback = class_config_llog_handler; /* set up client obds */ err = lustre_process_log(sb, profilenm, cfg); if (err < 0) { @@ -997,8 +1044,8 @@ out_free: OBD_FREE(dt, strlen(lprof->lp_dt) + instlen + 2); if (err) ll_put_super(sb); - else - LCONSOLE_WARN("Client %s has started\n", profilenm); + else if (sbi->ll_flags & LL_SBI_VERBOSE) + LCONSOLE_WARN("Mounted %s\n", profilenm); OBD_FREE_PTR(cfg); RETURN(err); @@ -1050,6 +1097,9 @@ void ll_put_super(struct super_block *sb) class_manual_cleanup(obd); } + if (sbi->ll_flags & LL_SBI_VERBOSE) + LCONSOLE_WARN("Unmounted %s\n", profilenm ? profilenm : ""); + if (profilenm) class_del_profile(profilenm); @@ -1065,8 +1115,6 @@ void ll_put_super(struct super_block *sb) cl_env_cache_purge(~0); - LCONSOLE_WARN("client %p umount complete\n", cfg.cfg_instance); - cfs_module_put(THIS_MODULE); EXIT; @@ -1169,14 +1217,9 @@ void ll_clear_inode(struct inode *inode) * cl_object still uses inode lsm. */ cl_inode_fini(inode); + lli->lli_has_smd = false; - if (lli->lli_smd) { - obd_free_memmd(sbi->ll_dt_exp, &lli->lli_smd); - lli->lli_smd = NULL; - } - - - EXIT; + EXIT; } int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data, @@ -1196,24 +1239,24 @@ int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data, rc = md_setattr(sbi->ll_md_exp, op_data, NULL, 0, NULL, 0, &request, mod); - if (rc) { - ptlrpc_req_finished(request); - if (rc == -ENOENT) { - inode->i_nlink = 0; - /* Unlinked special device node? Or just a race? - * Pretend we done everything. */ - if (!S_ISREG(inode->i_mode) && - !S_ISDIR(inode->i_mode)) { - ia_valid = op_data->op_attr.ia_valid; - op_data->op_attr.ia_valid &= ~TIMES_SET_FLAGS; - rc = simple_setattr(dentry, &op_data->op_attr); - op_data->op_attr.ia_valid = ia_valid; - } - } else if (rc != -EPERM && rc != -EACCES && rc != -ETXTBSY) { - CERROR("md_setattr fails: rc = %d\n", rc); - } - RETURN(rc); - } + if (rc) { + ptlrpc_req_finished(request); + if (rc == -ENOENT) { + clear_nlink(inode); + /* Unlinked special device node? Or just a race? + * Pretend we done everything. */ + if (!S_ISREG(inode->i_mode) && + !S_ISDIR(inode->i_mode)) { + ia_valid = op_data->op_attr.ia_valid; + op_data->op_attr.ia_valid &= ~TIMES_SET_FLAGS; + rc = simple_setattr(dentry, &op_data->op_attr); + op_data->op_attr.ia_valid = ia_valid; + } + } else if (rc != -EPERM && rc != -EACCES && rc != -ETXTBSY) { + CERROR("md_setattr fails: rc = %d\n", rc); + } + RETURN(rc); + } rc = md_get_lustre_md(sbi->ll_md_exp, request, sbi->ll_dt_exp, sbi->ll_md_exp, &md); @@ -1295,6 +1338,50 @@ static int ll_setattr_ost(struct inode *inode, struct iattr *attr) return rc; } +#ifndef HAVE_VFS_INODE_NEWSIZE_OK +/** + * inode_newsize_ok - may this inode be truncated to a given size + * @inode: the inode to be truncated + * @offset: the new size to assign to the inode + * @Returns: 0 on success, -ve errno on failure + * + * inode_newsize_ok will check filesystem limits and ulimits to check that the + * new inode size is within limits. inode_newsize_ok will also send SIGXFSZ + * when necessary. Caller must not proceed with inode size change if failure is + * returned. @inode must be a file (not directory), with appropriate + * permissions to allow truncate (inode_newsize_ok does NOT check these + * conditions). + * + * inode_newsize_ok must be called with i_mutex held. + */ +int inode_newsize_ok(const struct inode *inode, loff_t offset) +{ + if (inode->i_size < offset) { + unsigned long limit; + + limit = rlimit(RLIMIT_FSIZE); + if (limit != RLIM_INFINITY && offset > limit) + goto out_sig; + if (offset > inode->i_sb->s_maxbytes) + goto out_big; + } else { + /* + * truncation of in-use swapfiles is disallowed - it would + * cause subsequent swapout to scribble on the now-freed + * blocks. + */ + if (IS_SWAPFILE(inode)) + return -ETXTBSY; + } + + return 0; +out_sig: + send_sig(SIGXFSZ, current, 0); +out_big: + return -EFBIG; +} +#endif + /* If this inode has objects allocated to it (lsm != NULL), then the OST * object(s) determine the file size and mtime. Otherwise, the MDS will * keep these values until such a time that objects are allocated for it. @@ -1310,17 +1397,19 @@ static int ll_setattr_ost(struct inode *inode, struct iattr *attr) */ int ll_setattr_raw(struct dentry *dentry, struct iattr *attr) { + struct lov_stripe_md *lsm; 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; struct md_open_data *mod = NULL; int ia_valid = attr->ia_valid; int rc = 0, rc1 = 0; ENTRY; - CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu valid %x\n", inode->i_ino, - attr->ia_valid); + CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu (%p) to %llu, valid %x\n", + inode->i_ino, + inode,i_size_read(inode), + attr->ia_valid); if (ia_valid & ATTR_SIZE) { /* Check new size against VFS/VM file size limit and rlimit */ @@ -1367,11 +1456,6 @@ int ll_setattr_raw(struct dentry *dentry, struct iattr *attr) LTIME_S(attr->ia_mtime), LTIME_S(attr->ia_ctime), cfs_time_current_sec()); - /* NB: ATTR_SIZE will only be set after this point if the size - * resides on the MDS, ie, this file has no objects. */ - if (lsm) - attr->ia_valid &= ~ATTR_SIZE; - /* We always do an MDS RPC, even if we're only changing the size; * only the MDS knows whether truncate() should fail with -ETXTBUSY */ @@ -1379,15 +1463,24 @@ int ll_setattr_raw(struct dentry *dentry, struct iattr *attr) if (op_data == NULL) RETURN(-ENOMEM); - if (!S_ISDIR(inode->i_mode)) { - if (ia_valid & ATTR_SIZE) - UP_WRITE_I_ALLOC_SEM(inode); - UNLOCK_INODE_MUTEX(inode); - cfs_down_write(&lli->lli_trunc_sem); - LOCK_INODE_MUTEX(inode); - if (ia_valid & ATTR_SIZE) - DOWN_WRITE_I_ALLOC_SEM(inode); - } + if (!S_ISDIR(inode->i_mode)) { + if (ia_valid & ATTR_SIZE) + inode_dio_write_done(inode); + mutex_unlock(&inode->i_mutex); + cfs_down_write(&lli->lli_trunc_sem); + mutex_lock(&inode->i_mutex); + if (ia_valid & ATTR_SIZE) + inode_dio_wait(inode); + } + + /* We need a steady stripe configuration for setattr to avoid + * confusion. */ + lsm = ccc_inode_lsm_get(inode); + + /* NB: ATTR_SIZE will only be set after this point if the size + * resides on the MDS, ie, this file has no objects. */ + if (lsm != NULL) + attr->ia_valid &= ~ATTR_SIZE; memcpy(&op_data->op_attr, attr, sizeof(*attr)); @@ -1401,7 +1494,7 @@ int ll_setattr_raw(struct dentry *dentry, struct iattr *attr) GOTO(out, rc); ll_ioepoch_open(lli, op_data->op_ioepoch); - if (!lsm || !S_ISREG(inode->i_mode)) { + if (lsm == NULL || !S_ISREG(inode->i_mode)) { CDEBUG(D_INODE, "no lsm: not setting attrs on OST\n"); GOTO(out, rc = 0); } @@ -1420,6 +1513,7 @@ int ll_setattr_raw(struct dentry *dentry, struct iattr *attr) rc = ll_setattr_ost(inode, attr); EXIT; out: + ccc_inode_lsm_put(inode, lsm); if (op_data) { if (op_data->op_ioepoch) { rc1 = ll_setattr_done_writing(inode, op_data, mod); @@ -1473,7 +1567,7 @@ int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs, int rc; ENTRY; - rc = obd_statfs(class_exp2obd(sbi->ll_md_exp), osfs, max_age, flags); + rc = obd_statfs(NULL, sbi->ll_md_exp, osfs, max_age, flags); if (rc) { CERROR("md_statfs fails: rc = %d\n", rc); RETURN(rc); @@ -1487,8 +1581,7 @@ int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs, if (sbi->ll_flags & LL_SBI_LAZYSTATFS) flags |= OBD_STATFS_NODELAY; - rc = obd_statfs_rqset(class_exp2obd(sbi->ll_dt_exp), - &obd_osfs, max_age, flags); + rc = obd_statfs_rqset(sbi->ll_dt_exp, &obd_osfs, max_age, flags); if (rc) { CERROR("obd_statfs fails: rc = %d\n", rc); RETURN(rc); @@ -1515,16 +1608,11 @@ int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs, RETURN(rc); } -#ifndef HAVE_STATFS_DENTRY_PARAM -int ll_statfs(struct super_block *sb, struct kstatfs *sfs) -{ -#else int ll_statfs(struct dentry *de, struct kstatfs *sfs) { - struct super_block *sb = de->d_sb; -#endif - struct obd_statfs osfs; - int rc; + struct super_block *sb = de->d_sb; + struct obd_statfs osfs; + int rc; CDEBUG(D_VFSTRACE, "VFS Op: at "LPU64" jiffies\n", get_jiffies_64()); ll_stats_ops_tally(ll_s2sbi(sb), LPROC_LL_STAFS, 1); @@ -1559,10 +1647,9 @@ int ll_statfs(struct dentry *de, struct kstatfs *sfs) return 0; } -void ll_inode_size_lock(struct inode *inode, int lock_lsm) +void ll_inode_size_lock(struct inode *inode) { struct ll_inode_info *lli; - struct lov_stripe_md *lsm; LASSERT(!S_ISDIR(inode->i_mode)); @@ -1571,24 +1658,13 @@ void ll_inode_size_lock(struct inode *inode, int lock_lsm) cfs_down(&lli->lli_size_sem); LASSERT(lli->lli_size_sem_owner == NULL); lli->lli_size_sem_owner = current; - lsm = lli->lli_smd; - LASSERTF(lsm != NULL || lock_lsm == 0, "lsm %p, lock_lsm %d\n", - lsm, lock_lsm); - if (lock_lsm) - lov_stripe_lock(lsm); } -void ll_inode_size_unlock(struct inode *inode, int unlock_lsm) +void ll_inode_size_unlock(struct inode *inode) { struct ll_inode_info *lli; - struct lov_stripe_md *lsm; lli = ll_i2info(inode); - lsm = lli->lli_smd; - LASSERTF(lsm != NULL || unlock_lsm == 0, "lsm %p, lock_lsm %d\n", - lsm, unlock_lsm); - if (unlock_lsm) - lov_stripe_unlock(lsm); LASSERT(lli->lli_size_sem_owner == current); lli->lli_size_sem_owner = NULL; cfs_up(&lli->lli_size_sem); @@ -1596,54 +1672,31 @@ void ll_inode_size_unlock(struct inode *inode, int unlock_lsm) void ll_update_inode(struct inode *inode, struct lustre_md *md) { - struct ll_inode_info *lli = ll_i2info(inode); - struct mdt_body *body = md->body; - struct lov_stripe_md *lsm = md->lsm; - struct ll_sb_info *sbi = ll_i2sbi(inode); - - LASSERT ((lsm != NULL) == ((body->valid & OBD_MD_FLEASIZE) != 0)); - if (lsm != NULL) { - LASSERT(S_ISREG(inode->i_mode)); - - cfs_mutex_lock(&lli->lli_och_mutex); - if (lli->lli_smd == NULL) { - if (lsm->lsm_magic != LOV_MAGIC_V1 && - lsm->lsm_magic != LOV_MAGIC_V3) { - dump_lsm(D_ERROR, lsm); - LBUG(); - } - CDEBUG(D_INODE, "adding lsm %p to inode %lu/%u(%p)\n", - lsm, inode->i_ino, inode->i_generation, inode); - /* 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_file_inode_init(inode, md); - cfs_spin_lock(&lli->lli_lock); - lli->lli_smd = lsm; - cfs_spin_unlock(&lli->lli_lock); - cfs_mutex_unlock(&lli->lli_och_mutex); - lli->lli_maxbytes = lsm->lsm_maxbytes; - if (lli->lli_maxbytes > MAX_LFS_FILESIZE) - lli->lli_maxbytes = MAX_LFS_FILESIZE; - } else { - cfs_mutex_unlock(&lli->lli_och_mutex); - LASSERT(lli->lli_smd->lsm_magic == lsm->lsm_magic && - lli->lli_smd->lsm_stripe_count == - lsm->lsm_stripe_count); - if (lov_stripe_md_cmp(lli->lli_smd, lsm)) { - CERROR("lsm mismatch for inode %ld\n", - inode->i_ino); - CERROR("lli_smd:\n"); - dump_lsm(D_ERROR, lli->lli_smd); - CERROR("lsm:\n"); - dump_lsm(D_ERROR, lsm); - LBUG(); - } - } - if (lli->lli_smd != lsm) - obd_free_memmd(ll_i2dtexp(inode), &lsm); - } + struct ll_inode_info *lli = ll_i2info(inode); + struct mdt_body *body = md->body; + struct lov_stripe_md *lsm = md->lsm; + struct ll_sb_info *sbi = ll_i2sbi(inode); + + LASSERT ((lsm != NULL) == ((body->valid & OBD_MD_FLEASIZE) != 0)); + if (lsm != NULL) { + LASSERT(S_ISREG(inode->i_mode)); + cfs_mutex_lock(&lli->lli_och_mutex); + CDEBUG(D_INODE, "adding lsm %p to inode %lu/%u(%p)\n", + lsm, inode->i_ino, inode->i_generation, inode); + /* cl_file_inode_init must go before lli_has_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 */ + if (cl_file_inode_init(inode, md) == 0) + lli->lli_has_smd = true; + cfs_mutex_unlock(&lli->lli_och_mutex); + + lli->lli_maxbytes = lsm->lsm_maxbytes; + if (lli->lli_maxbytes > MAX_LFS_FILESIZE) + lli->lli_maxbytes = MAX_LFS_FILESIZE; + if (md->lsm != NULL) + obd_free_memmd(ll_i2dtexp(inode), &md->lsm); + } if (sbi->ll_flags & LL_SBI_RMT_CLIENT) { if (body->valid & OBD_MD_FLRMTPERM) @@ -1690,19 +1743,16 @@ void ll_update_inode(struct inode *inode, struct lustre_md *md) } else { inode->i_blkbits = inode->i_sb->s_blocksize_bits; } -#ifdef HAVE_INODE_BLKSIZE - inode->i_blksize = 1<i_blkbits; -#endif - if (body->valid & OBD_MD_FLUID) - inode->i_uid = body->uid; - if (body->valid & OBD_MD_FLGID) - inode->i_gid = body->gid; - if (body->valid & OBD_MD_FLFLAGS) - inode->i_flags = ll_ext_to_inode_flags(body->flags); - if (body->valid & OBD_MD_FLNLINK) - inode->i_nlink = body->nlink; - if (body->valid & OBD_MD_FLRDEV) - inode->i_rdev = old_decode_dev(body->rdev); + if (body->valid & OBD_MD_FLUID) + inode->i_uid = body->uid; + if (body->valid & OBD_MD_FLGID) + inode->i_gid = body->gid; + if (body->valid & OBD_MD_FLFLAGS) + inode->i_flags = ll_ext_to_inode_flags(body->flags); + if (body->valid & OBD_MD_FLNLINK) + set_nlink(inode, body->nlink); + if (body->valid & OBD_MD_FLRDEV) + inode->i_rdev = old_decode_dev(body->rdev); if (body->valid & OBD_MD_FLID) { /* FID shouldn't be changed! */ @@ -1720,7 +1770,7 @@ void ll_update_inode(struct inode *inode, struct lustre_md *md) if (body->valid & OBD_MD_FLSIZE) { if (exp_connect_som(ll_i2mdexp(inode)) && - S_ISREG(inode->i_mode) && lli->lli_smd) { + S_ISREG(inode->i_mode) && lli->lli_has_smd) { struct lustre_handle lockh; ldlm_mode_t mode; @@ -1778,7 +1828,7 @@ void ll_read_inode2(struct inode *inode, void *opaque) CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p)\n", PFID(&lli->lli_fid), inode); - LASSERT(!lli->lli_smd); + LASSERT(!lli->lli_has_smd); /* Core attributes from the MDS first. This is a new inode, and * the VFS doesn't zero times in the core inode so we have to do @@ -1805,7 +1855,6 @@ void ll_read_inode2(struct inode *inode, void *opaque) } 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 = (struct address_space_operations *)&ll_dir_aops; EXIT; } else if (S_ISLNK(inode->i_mode)) { inode->i_op = &ll_fast_symlink_inode_operations; @@ -1813,8 +1862,8 @@ void ll_read_inode2(struct inode *inode, void *opaque) } else { inode->i_op = &ll_special_inode_operations; - init_special_inode(inode, inode->i_mode, - kdev_t_to_nr(inode->i_rdev)); + init_special_inode(inode, inode->i_mode, + inode->i_rdev); EXIT; } @@ -1822,13 +1871,19 @@ void ll_read_inode2(struct inode *inode, void *opaque) void ll_delete_inode(struct inode *inode) { - struct ll_sb_info *sbi = ll_i2sbi(inode); - int rc; - ENTRY; + struct cl_inode_info *lli = cl_i2info(inode); + struct ll_sb_info *sbi = ll_i2sbi(inode); + int rc; + ENTRY; - rc = obd_fid_delete(sbi->ll_md_exp, ll_inode2fid(inode)); - if (rc) - CERROR("fid_delete() failed, rc %d\n", rc); + rc = obd_fid_delete(sbi->ll_md_exp, ll_inode2fid(inode)); + if (rc) + CERROR("fid_delete() failed, rc %d\n", rc); + + if (S_ISREG(inode->i_mode) && lli->lli_clob != NULL) + /* discard all dirty pages before truncating them, required by + * osc_extent implementation at LU-1030. */ + cl_sync_file_range(inode, 0, OBD_OBJECT_EOF, CL_FSYNC_DISCARD); truncate_inode_pages(&inode->i_data, 0); @@ -1890,18 +1945,13 @@ int ll_iocontrol(struct inode *inode, struct file *file, RETURN(put_user(flags, (int *)arg)); } case FSFILT_IOC_SETFLAGS: { - struct lov_stripe_md *lsm = ll_i2info(inode)->lli_smd; + struct lov_stripe_md *lsm; struct obd_info oinfo = { { { 0 } } }; struct md_op_data *op_data; if (get_user(flags, (int *)arg)) RETURN(-EFAULT); - oinfo.oi_md = lsm; - OBDO_ALLOC(oinfo.oi_oa); - if (!oinfo.oi_oa) - RETURN(-ENOMEM); - op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL, 0, 0, LUSTRE_OPC_ANY, NULL); if (IS_ERR(op_data)) @@ -1913,37 +1963,37 @@ int ll_iocontrol(struct inode *inode, struct file *file, NULL, 0, NULL, 0, &req, NULL); ll_finish_md_op_data(op_data); ptlrpc_req_finished(req); - if (rc) { - OBDO_FREE(oinfo.oi_oa); - RETURN(rc); - } + if (rc) + RETURN(rc); - if (lsm == NULL) { - OBDO_FREE(oinfo.oi_oa); - GOTO(update_cache, rc); - } + inode->i_flags = ll_ext_to_inode_flags(flags); + lsm = ccc_inode_lsm_get(inode); + if (lsm == NULL) + RETURN(0); + + OBDO_ALLOC(oinfo.oi_oa); + if (!oinfo.oi_oa) { + ccc_inode_lsm_put(inode, lsm); + RETURN(-ENOMEM); + } + oinfo.oi_md = lsm; oinfo.oi_oa->o_id = lsm->lsm_object_id; 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, - &ll_i2info(inode)->lli_fid, 0); + obdo_set_parent_fid(oinfo.oi_oa, &ll_i2info(inode)->lli_fid); rc = obd_setattr_rqset(sbi->ll_dt_exp, &oinfo, NULL); capa_put(oinfo.oi_capa); OBDO_FREE(oinfo.oi_oa); - if (rc) { - if (rc != -EPERM && rc != -EACCES) - CERROR("osc_setattr_async fails: rc = %d\n",rc); - RETURN(rc); - } + ccc_inode_lsm_put(inode, lsm); - EXIT; -update_cache: - inode->i_flags = ll_ext_to_inode_flags(flags); - return 0; + if (rc && rc != -EPERM && rc != -EACCES) + CERROR("osc_setattr_async fails: rc = %d\n", rc); + + RETURN(rc); } default: RETURN(-ENOSYS); @@ -1958,10 +2008,10 @@ int ll_flush_ctx(struct inode *inode) CDEBUG(D_SEC, "flush context for user %d\n", cfs_curproc_uid()); - obd_set_info_async(sbi->ll_md_exp, + obd_set_info_async(NULL, sbi->ll_md_exp, sizeof(KEY_FLUSH_CTX), KEY_FLUSH_CTX, 0, NULL, NULL); - obd_set_info_async(sbi->ll_dt_exp, + obd_set_info_async(NULL, sbi->ll_dt_exp, sizeof(KEY_FLUSH_CTX), KEY_FLUSH_CTX, 0, NULL, NULL); return 0; @@ -1976,7 +2026,6 @@ void ll_umount_begin(struct vfsmount *vfsmnt, int flags) void ll_umount_begin(struct super_block *sb) { #endif - struct lustre_sb_info *lsi = s2lsi(sb); struct ll_sb_info *sbi = ll_s2sbi(sb); struct obd_device *obd; struct obd_ioctl_data *ioc_data; @@ -1989,9 +2038,6 @@ void ll_umount_begin(struct super_block *sb) } #endif - /* Tell the MGC we got umount -f */ - lsi->lsi_flags |= LSI_UMOUNT_FORCE; - CDEBUG(D_VFSTRACE, "VFS Op: superblock %p count %d active %d\n", sb, sb->s_count, atomic_read(&sb->s_active)); @@ -2047,18 +2093,20 @@ void ll_umount_begin(struct super_block *sb) int ll_remount_fs(struct super_block *sb, int *flags, char *data) { struct ll_sb_info *sbi = ll_s2sbi(sb); + char *profilenm = get_profile_name(sb); int err; __u32 read_only; if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) { read_only = *flags & MS_RDONLY; - err = obd_set_info_async(sbi->ll_md_exp, + err = obd_set_info_async(NULL, sbi->ll_md_exp, sizeof(KEY_READ_ONLY), KEY_READ_ONLY, sizeof(read_only), &read_only, NULL); if (err) { - CERROR("Failed to change the read-only flag during " - "remount: %d\n", err); + LCONSOLE_WARN("Failed to remount %s %s (%d)\n", + profilenm, read_only ? + "read-only" : "read-write", err); return err; } @@ -2066,6 +2114,10 @@ int ll_remount_fs(struct super_block *sb, int *flags, char *data) sb->s_flags |= MS_RDONLY; else sb->s_flags &= ~MS_RDONLY; + + if (sbi->ll_flags & LL_SBI_VERBOSE) + LCONSOLE_WARN("Remounted %s %s\n", profilenm, + read_only ? "read-only" : "read-write"); } return 0; } @@ -2074,8 +2126,9 @@ int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req, struct super_block *sb) { - struct ll_sb_info *sbi = NULL; - struct lustre_md md; + struct ll_sb_info *sbi = NULL; + struct lustre_md md; + __u64 ibits; int rc; ENTRY; @@ -2114,9 +2167,18 @@ int ll_prep_inode(struct inode **inode, } } + /* sanity check for LAYOUT lock. */ + ibits = MDS_INODELOCK_LAYOUT; + if (S_ISREG(md.body->mode) && sbi->ll_flags & LL_SBI_LAYOUT_LOCK && + md.lsm != NULL && !ll_have_md_lock(*inode, &ibits, LCK_MINMODE)) { + CERROR("%s: inode "DFID" (%p) layout lock not granted.\n", + ll_get_fsname(sb, NULL, 0), + PFID(ll_inode2fid(*inode)), *inode); + } + out: - md_free_lustre_md(sbi->ll_md_exp, &md); - RETURN(rc); + md_free_lustre_md(sbi->ll_md_exp, &md); + RETURN(rc); } int ll_obd_statfs(struct inode *inode, void *arg) @@ -2126,6 +2188,7 @@ int ll_obd_statfs(struct inode *inode, void *arg) char *buf = NULL; struct obd_ioctl_data *data = NULL; __u32 type; + __u32 flags; int len = 0, rc; if (!inode || !(sbi = ll_i2sbi(inode))) @@ -2147,14 +2210,15 @@ 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_LMV) + if (type & LL_STATFS_LMV) exp = sbi->ll_md_exp; - else if (type == LL_STATFS_LOV) + else if (type & LL_STATFS_LOV) exp = sbi->ll_dt_exp; else GOTO(out_statfs, rc = -ENODEV); - rc = obd_iocontrol(IOC_OBD_STATFS, exp, len, buf, NULL); + flags = (type & LL_STATFS_NODELAY) ? OBD_STATFS_NODELAY : 0; + rc = obd_iocontrol(IOC_OBD_STATFS, exp, len, buf, &flags); if (rc) GOTO(out_statfs, rc); out_statfs: @@ -2237,12 +2301,11 @@ struct md_op_data * ll_prep_md_op_data(struct md_op_data *op_data, * 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); + !ll_i2info(i2)->lli_has_smd) { + 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))) + cfs_spin_lock(&lli->lli_lock); + if (likely(!lli->lli_has_smd && !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. */ @@ -2280,6 +2343,9 @@ int ll_show_options(struct seq_file *seq, struct vfsmount *vfs) if (sbi->ll_flags & LL_SBI_LAZYSTATFS) seq_puts(seq, ",lazystatfs"); + if (sbi->ll_flags & LL_SBI_USER_FID2PATH) + seq_puts(seq, ",user_fid2path"); + RETURN(0); } @@ -2308,3 +2374,81 @@ int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg) RETURN(0); } + +/** + * Get lustre file system name by \a sbi. If \a buf is provided(non-NULL), the + * fsname will be returned in this buffer; otherwise, a static buffer will be + * used to store the fsname and returned to caller. + */ +char *ll_get_fsname(struct super_block *sb, char *buf, int buflen) +{ + static char fsname_static[MTI_NAME_MAXLEN]; + struct lustre_sb_info *lsi = s2lsi(sb); + char *ptr; + int len; + + if (buf == NULL) { + /* this means the caller wants to use static buffer + * and it doesn't care about race. Usually this is + * in error reporting path */ + buf = fsname_static; + buflen = sizeof(fsname_static); + } + + len = strlen(lsi->lsi_lmd->lmd_profile); + ptr = strrchr(lsi->lsi_lmd->lmd_profile, '-'); + if (ptr && (strcmp(ptr, "-client") == 0)) + len -= 7; + + if (unlikely(len >= buflen)) + len = buflen - 1; + strncpy(buf, lsi->lsi_lmd->lmd_profile, len); + buf[len] = '\0'; + + return buf; +} + +static char* ll_d_path(struct dentry *dentry, char *buf, int bufsize) +{ + char *path = NULL; + +#ifdef HAVE_FS_STRUCT_USE_PATH + struct path p; + + p.dentry = dentry; + p.mnt = current->fs->root.mnt; + path_get(&p); + path = d_path(&p, buf, bufsize); + path_put(&p); +#else + path = d_path(dentry, current->fs->rootmnt, buf, bufsize); +#endif + + return path; +} + +void ll_dirty_page_discard_warn(cfs_page_t *page, int ioret) +{ + char *buf, *path = NULL; + struct dentry *dentry = NULL; + struct ccc_object *obj = cl_inode2ccc(page->mapping->host); + + buf = (char *)__get_free_page(GFP_KERNEL); + if (buf != NULL) { + dentry = d_find_alias(page->mapping->host); + if (dentry != NULL) + path = ll_d_path(dentry, buf, PAGE_SIZE); + } + + CWARN("%s: dirty page discard: %s/fid: "DFID"/%s may get corrupted " + "(rc %d)\n", ll_get_fsname(page->mapping->host->i_sb, NULL, 0), + s2lsi(page->mapping->host->i_sb)->lsi_lmd->lmd_dev, + PFID(&obj->cob_header.coh_lu.loh_fid), + (path && !IS_ERR(path)) ? path : "", ioret); + + if (dentry != NULL) + dput(dentry); + + if (buf != NULL) + free_page((unsigned long)buf); +}