X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fllite%2Fllite_lib.c;h=efaa76fc9b741e211cbe09aa9a69a3c7ec89d1ee;hp=83a131eb3f86d8cf18dfa92a9f68d51ef4159175;hb=51206e8cd42134400fa0b6259a92d7138f3dc984;hpb=138544d2c4b5332902d8141c0d119074123e072b diff --git a/lustre/llite/llite_lib.c b/lustre/llite/llite_lib.c index 83a131e..efaa76f 100644 --- a/lustre/llite/llite_lib.c +++ b/lustre/llite/llite_lib.c @@ -26,8 +26,10 @@ * 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, 2012, Whamcloud, Inc. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -42,7 +44,6 @@ #include #include -#include #include #include @@ -59,14 +60,19 @@ cfs_mem_cache_t *ll_file_data_slab; -LIST_HEAD(ll_super_blocks); -spinlock_t ll_sb_lock = SPIN_LOCK_UNLOCKED; +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) ffz(~(n)) +#define log2(n) cfs_ffz(~(n)) #endif static struct ll_sb_info *ll_init_sbi(void) @@ -82,10 +88,10 @@ static struct ll_sb_info *ll_init_sbi(void) if (!sbi) RETURN(NULL); - spin_lock_init(&sbi->ll_lock); - spin_lock_init(&sbi->ll_lco.lco_lock); - spin_lock_init(&sbi->ll_pp_extent_lock); - spin_lock_init(&sbi->ll_process_lock); + cfs_spin_lock_init(&sbi->ll_lock); + cfs_mutex_init(&sbi->ll_lco.lco_lock); + cfs_spin_lock_init(&sbi->ll_pp_extent_lock); + cfs_spin_lock_init(&sbi->ll_process_lock); sbi->ll_rw_stats_on = 0; si_meminfo(&si); @@ -105,18 +111,19 @@ static struct ll_sb_info *ll_init_sbi(void) sbi->ll_ra_info.ra_max_pages = sbi->ll_ra_info.ra_max_pages_per_file; sbi->ll_ra_info.ra_max_read_ahead_whole_pages = SBI_DEFAULT_READAHEAD_WHOLE_MAX; - INIT_LIST_HEAD(&sbi->ll_conn_chain); - INIT_LIST_HEAD(&sbi->ll_orphan_dentry_list); + CFS_INIT_LIST_HEAD(&sbi->ll_conn_chain); + CFS_INIT_LIST_HEAD(&sbi->ll_orphan_dentry_list); ll_generate_random_uuid(uuid); class_uuid_unparse(uuid, &sbi->ll_sb_uuid); CDEBUG(D_CONFIG, "generated uuid: %s\n", sbi->ll_sb_uuid.uuid); - spin_lock(&ll_sb_lock); - list_add_tail(&sbi->ll_list, &ll_super_blocks); - spin_unlock(&ll_sb_lock); + cfs_spin_lock(&ll_sb_lock); + cfs_list_add_tail(&sbi->ll_list, &ll_super_blocks); + cfs_spin_unlock(&ll_sb_lock); -#ifdef ENABLE_LLITE_CHECKSUM + sbi->ll_flags |= LL_SBI_VERBOSE; +#ifdef ENABLE_CHECKSUM sbi->ll_flags |= LL_SBI_CHECKSUM; #endif @@ -124,17 +131,19 @@ static struct ll_sb_info *ll_init_sbi(void) sbi->ll_flags |= LL_SBI_LRU_RESIZE; #endif -#ifdef HAVE_EXPORT___IGET - INIT_LIST_HEAD(&sbi->ll_deathrow); - spin_lock_init(&sbi->ll_deathrow_lock); -#endif for (i = 0; i <= LL_PROCESS_HIST_MAX; i++) { - spin_lock_init(&sbi->ll_rw_extents_info.pp_extents[i].pp_r_hist.oh_lock); - spin_lock_init(&sbi->ll_rw_extents_info.pp_extents[i].pp_w_hist.oh_lock); + cfs_spin_lock_init(&sbi->ll_rw_extents_info.pp_extents[i]. \ + pp_r_hist.oh_lock); + cfs_spin_lock_init(&sbi->ll_rw_extents_info.pp_extents[i]. \ + pp_w_hist.oh_lock); } /* metadata statahead is enabled by default */ sbi->ll_sa_max = LL_SA_RPC_DEF; + 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); } @@ -145,31 +154,31 @@ void ll_free_sbi(struct super_block *sb) ENTRY; if (sbi != NULL) { - spin_lock(&ll_sb_lock); - list_del(&sbi->ll_list); - spin_unlock(&ll_sb_lock); + cfs_spin_lock(&ll_sb_lock); + cfs_list_del(&sbi->ll_list); + cfs_spin_unlock(&ll_sb_lock); OBD_FREE(sbi, sizeof(*sbi)); } EXIT; } static struct dentry_operations ll_d_root_ops = { -#ifdef DCACHE_LUSTRE_INVALID .d_compare = ll_dcompare, -#endif .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; + struct md_op_data *op_data; struct lustre_md lmd; obd_valid valid; int size, err, checksum; @@ -185,6 +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); @@ -194,12 +209,16 @@ 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_SOM; + 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; #ifdef HAVE_LRU_RESIZE_SUPPORT if (sbi->ll_flags & LL_SBI_LRU_RESIZE) @@ -220,6 +239,9 @@ 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; @@ -233,6 +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 " @@ -252,10 +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, cfs_time_current_64() - 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); @@ -264,25 +313,12 @@ 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(): - * retval = in_file->f_op->sendfile(...); - * if (*ppos > max) - * retval = -EOVERFLOW; - * - * it will check if *ppos is greater than max. However, max equals to - * s_maxbytes, which is a negative integer in a x86_64 box since loff_t - * has been defined as a signed long long ineger in linux kernel. */ -#if BITS_PER_LONG == 64 - sb->s_maxbytes = PAGE_CACHE_MAXBYTES >> 1; -#else - sb->s_maxbytes = PAGE_CACHE_MAXBYTES; -#endif - sbi->ll_namelen = osfs.os_namelen; + sb->s_maxbytes = MAX_LFS_FILESIZE; + sbi->ll_namelen = osfs->os_namelen; sbi->ll_max_rw_chunk = LL_DEFAULT_MAX_RW_CHUNK; if ((sbi->ll_flags & LL_SBI_USER_XATTR) && @@ -305,9 +341,6 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) sbi->ll_flags &= ~LL_SBI_ACL; } - if (data->ocd_connect_flags & OBD_CONNECT_JOIN) - sbi->ll_flags |= LL_SBI_JOIN; - if (data->ocd_connect_flags & OBD_CONNECT_RMT_CLIENT) { if (!(sbi->ll_flags & LL_SBI_RMT_CLIENT)) { sbi->ll_flags |= LL_SBI_RMT_CLIENT; @@ -331,6 +364,14 @@ 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); @@ -343,7 +384,12 @@ 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_SOM; + OBD_CONNECT_FULL20 | OBD_CONNECT_64BITHASH | + OBD_CONNECT_MAXBYTES | + OBD_CONNECT_EINPROGRESS; + + if (sbi->ll_flags & LL_SBI_SOM_PREVIEW) + data->ocd_connect_flags |= OBD_CONNECT_SOM; if (!OBD_FAIL_CHECK(OBD_FAIL_OSC_CONNECT_CKSUM)) { /* OBD_CONNECT_CKSUM should always be set, even if checksums are @@ -355,8 +401,7 @@ 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 @@ -371,7 +416,8 @@ 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) { @@ -392,11 +438,11 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) GOTO(out_dt, err); } - spin_lock(&sbi->ll_lco.lco_lock); + cfs_mutex_lock(&sbi->ll_lco.lco_lock); sbi->ll_lco.lco_flags = data->ocd_connect_flags; sbi->ll_lco.lco_md_exp = sbi->ll_md_exp; sbi->ll_lco.lco_dt_exp = sbi->ll_dt_exp; - spin_unlock(&sbi->ll_lco.lco_lock); + cfs_mutex_unlock(&sbi->ll_lco.lco_lock); fid_zero(&sbi->ll_root_fid); err = md_getstatus(sbi->ll_md_exp, &sbi->ll_root_fid, &oc); @@ -411,7 +457,9 @@ 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? */ @@ -421,15 +469,23 @@ static int client_common_fill_super(struct super_block *sb, char *md, char *dt) else if (sbi->ll_flags & LL_SBI_ACL) valid |= OBD_MD_FLACL; - err = md_getattr(sbi->ll_md_exp, &sbi->ll_root_fid, oc, valid, 0, - &request); + OBD_ALLOC_PTR(op_data); + if (op_data == NULL) + GOTO(out_lock_cn_cb, err = -ENOMEM); + + op_data->op_fid1 = sbi->ll_root_fid; + op_data->op_mode = 0; + op_data->op_capa1 = oc; + op_data->op_valid = valid; + + err = md_getattr(sbi->ll_md_exp, op_data, &request); if (oc) capa_put(oc); + OBD_FREE_PTR(op_data); if (err) { 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) { @@ -439,7 +495,7 @@ 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, ll_fid_build_ino(sbi, &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); @@ -478,9 +534,6 @@ 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; @@ -493,6 +546,12 @@ 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: @@ -511,6 +570,8 @@ out_md: out: if (data != NULL) OBD_FREE_PTR(data); + if (osfs != NULL) + OBD_FREE_PTR(osfs); lprocfs_unregister_mountpoint(sbi); return err; } @@ -532,26 +593,16 @@ int ll_get_max_mdsize(struct ll_sb_info *sbi, int *lmmsize) void ll_dump_inode(struct inode *inode) { struct list_head *tmp; - struct dentry *dentry; int dentry_count = 0; LASSERT(inode != NULL); - CERROR("inode %p dump: dev=%s ino=%lu mode=%o count=%u state %lx\n", - inode, ll_i2mdexp(inode)->exp_obd->obd_name, inode->i_ino, - inode->i_mode, atomic_read(&inode->i_count), inode->i_state); - - list_for_each(tmp, &inode->i_dentry) { - dentry = list_entry(tmp, struct dentry, d_alias); - CERROR("Alias[%d] dentry %p dump: name=%.*s parent=%.*s (%p), inode=%p, count=%u, " - "flags=0x%x, fsdata=%p\n", dentry_count, 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_flags, dentry->d_fsdata); + list_for_each(tmp, &inode->i_dentry) dentry_count++; - } + CERROR("inode %p dump: dev=%s ino=%lu mode=%o count=%u, %d dentries\n", + inode, ll_i2mdexp(inode)->exp_obd->obd_name, inode->i_ino, + inode->i_mode, atomic_read(&inode->i_count), dentry_count); } void lustre_dump_dentry(struct dentry *dentry, int recur) @@ -576,109 +627,12 @@ void lustre_dump_dentry(struct dentry *dentry, int recur) if (recur == 0) return; - list_for_each(tmp, &dentry->d_subdirs) { + list_for_each(tmp, &dentry->d_subdirs) { struct dentry *d = list_entry(tmp, struct dentry, d_child); lustre_dump_dentry(d, recur - 1); } } -#ifdef HAVE_EXPORT___IGET -static void prune_dir_dentries(struct inode *inode) -{ - struct dentry *dentry, *prev = NULL; - - /* due to lustre specific logic, a directory - * can have few dentries - a bug from VFS POV */ -restart: - spin_lock(&dcache_lock); - if (!list_empty(&inode->i_dentry)) { - dentry = list_entry(inode->i_dentry.prev, - struct dentry, d_alias); - /* in order to prevent infinite loops we - * break if previous dentry is busy */ - if (dentry != prev) { - prev = dentry; - dget_locked(dentry); - spin_unlock(&dcache_lock); - - /* try to kill all child dentries */ - lock_dentry(dentry); - shrink_dcache_parent(dentry); - unlock_dentry(dentry); - dput(dentry); - - /* now try to get rid of current dentry */ - d_prune_aliases(inode); - goto restart; - } - } - spin_unlock(&dcache_lock); -} - -static void prune_deathrow_one(struct ll_inode_info *lli) -{ - struct inode *inode = ll_info2i(lli); - - /* first, try to drop any dentries - they hold a ref on the inode */ - if (S_ISDIR(inode->i_mode)) - prune_dir_dentries(inode); - else - d_prune_aliases(inode); - - - /* if somebody still uses it, leave it */ - LASSERT(atomic_read(&inode->i_count) > 0); - if (atomic_read(&inode->i_count) > 1) - goto out; - - CDEBUG(D_INODE, "inode %lu/%u(%d) looks a good candidate for prune\n", - inode->i_ino,inode->i_generation, atomic_read(&inode->i_count)); - - /* seems nobody uses it anymore */ - inode->i_nlink = 0; - -out: - iput(inode); - return; -} - -static void prune_deathrow(struct ll_sb_info *sbi, int try) -{ - struct ll_inode_info *lli; - int empty; - - do { - if (need_resched() && try) - break; - - if (try) { - if (!spin_trylock(&sbi->ll_deathrow_lock)) - break; - } else { - spin_lock(&sbi->ll_deathrow_lock); - } - - empty = 1; - lli = NULL; - if (!list_empty(&sbi->ll_deathrow)) { - lli = list_entry(sbi->ll_deathrow.next, - struct ll_inode_info, - lli_dead_list); - list_del_init(&lli->lli_dead_list); - if (!list_empty(&sbi->ll_deathrow)) - empty = 0; - } - spin_unlock(&sbi->ll_deathrow_lock); - - if (lli) - prune_deathrow_one(lli); - - } while (empty == 0); -} -#else /* !HAVE_EXPORT___IGET */ -#define prune_deathrow(sbi, try) do {} while (0) -#endif /* HAVE_EXPORT___IGET */ - void client_common_put_super(struct super_block *sb) { struct ll_sb_info *sbi = ll_s2sbi(sb); @@ -697,10 +651,7 @@ void client_common_put_super(struct super_block *sb) cl_sb_fini(sb); - /* destroy inodes in deathrow */ - prune_deathrow(sbi, 0); - - list_del(&sbi->ll_conn_chain); + cfs_list_del(&sbi->ll_conn_chain); obd_fid_fini(sbi->ll_dt_exp); obd_disconnect(sbi->ll_dt_exp); @@ -810,17 +761,25 @@ 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; @@ -857,7 +816,26 @@ static int ll_options(char *options, int *flags) *flags &= ~tmp; goto next; } - + tmp = ll_set_opt("som_preview", s1, LL_SBI_SOM_PREVIEW); + if (tmp) { + *flags |= tmp; + goto next; + } + tmp = ll_set_opt("32bitapi", s1, LL_SBI_32BIT_API); + if (tmp) { + *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); @@ -875,45 +853,97 @@ next: void ll_lli_init(struct ll_inode_info *lli) { lli->lli_inode_magic = LLI_INODE_MAGIC; - sema_init(&lli->lli_size_sem, 1); - sema_init(&lli->lli_write_sem, 1); lli->lli_flags = 0; - lli->lli_maxbytes = PAGE_CACHE_MAXBYTES; - spin_lock_init(&lli->lli_lock); - INIT_LIST_HEAD(&lli->lli_close_list); - lli->lli_inode_magic = LLI_INODE_MAGIC; - sema_init(&lli->lli_och_sem, 1); - lli->lli_mds_read_och = lli->lli_mds_write_och = NULL; + lli->lli_ioepoch = 0; + lli->lli_maxbytes = MAX_LFS_FILESIZE; + cfs_spin_lock_init(&lli->lli_lock); + lli->lli_posix_acl = NULL; + lli->lli_remote_perms = NULL; + cfs_mutex_init(&lli->lli_rmtperm_mutex); + /* 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_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 = lli->lli_open_fd_write_count = 0; + lli->lli_open_fd_read_count = 0; + lli->lli_open_fd_write_count = 0; lli->lli_open_fd_exec_count = 0; - INIT_LIST_HEAD(&lli->lli_dead_list); - lli->lli_remote_perms = NULL; - lli->lli_rmtperm_utime = 0; - sema_init(&lli->lli_rmtperm_sem, 1); - INIT_LIST_HEAD(&lli->lli_oss_capas); + cfs_mutex_init(&lli->lli_och_mutex); + 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_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; + } else { + cfs_sema_init(&lli->lli_size_sem, 1); + lli->lli_size_sem_owner = NULL; + 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; + } } -int ll_fill_super(struct super_block *sb) +static inline int ll_bdi_register(struct backing_dev_info *bdi) { - struct lustre_profile *lprof; +#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 + return 0; +#endif +} + +int ll_fill_super(struct super_block *sb, struct vfsmount *mnt) +{ + 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(); + cfs_module_put(THIS_MODULE); + OBD_FREE_PTR(cfg); RETURN(-ENOMEM); } @@ -921,20 +951,33 @@ 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" @@ -945,31 +988,30 @@ 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); + else if (sbi->ll_flags & LL_SBI_VERBOSE) + LCONSOLE_WARN("Mounted %s\n", profilenm); + OBD_FREE_PTR(cfg); RETURN(err); } /* ll_fill_super */ @@ -979,7 +1021,6 @@ 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); @@ -991,9 +1032,8 @@ 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); @@ -1021,9 +1061,17 @@ 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); + 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; @@ -1031,9 +1079,7 @@ void ll_put_super(struct super_block *sb) cl_env_cache_purge(~0); - LCONSOLE_WARN("client %s umount complete\n", ll_instance); - - cfs_module_put(); + cfs_module_put(THIS_MODULE); EXIT; } /* client_put_super */ @@ -1049,11 +1095,10 @@ 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; } } @@ -1084,8 +1129,8 @@ 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); } @@ -1104,7 +1149,7 @@ 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; @@ -1119,7 +1164,7 @@ void ll_clear_inode(struct inode *inode) } #ifdef CONFIG_FS_POSIX_ACL else if (lli->lli_posix_acl) { - LASSERT(atomic_read(&lli->lli_posix_acl->a_refcount) == 1); + LASSERT(cfs_atomic_read(&lli->lli_posix_acl->a_refcount) == 1); LASSERT(lli->lli_remote_perms == NULL); posix_acl_release(lli->lli_posix_acl); lli->lli_posix_acl = NULL; @@ -1127,12 +1172,10 @@ void ll_clear_inode(struct inode *inode) #endif lli->lli_inode_magic = LLI_INODE_DEAD; -#ifdef HAVE_EXPORT___IGET - spin_lock(&sbi->ll_deathrow_lock); - list_del_init(&lli->lli_dead_list); - spin_unlock(&sbi->ll_deathrow_lock); -#endif 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. @@ -1148,13 +1191,14 @@ 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; + int rc, ia_valid; ENTRY; op_data = ll_prep_md_op_data(op_data, inode, NULL, NULL, 0, 0, @@ -1171,8 +1215,12 @@ int ll_md_setattr(struct inode *inode, struct md_op_data *op_data, /* Unlinked special device node? Or just a race? * Pretend we done everything. */ if (!S_ISREG(inode->i_mode) && - !S_ISDIR(inode->i_mode)) - rc = inode_setattr(inode, &op_data->op_attr); + !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); } @@ -1186,12 +1234,16 @@ int ll_md_setattr(struct inode *inode, struct md_op_data *op_data, RETURN(rc); } - /* We call inode_setattr to adjust timestamps. + /* We want to adjust timestamps. * If there is at least some data in file, we cleared ATTR_SIZE - * above to avoid invoking vmtruncate, otherwise it is important - * to call vmtruncate in inode_setattr to update inode->i_size + * to avoid update size, otherwise it is important to do.(SOM case) * (bug 6196) */ - rc = inode_setattr(inode, &op_data->op_attr); + ia_valid = op_data->op_attr.ia_valid; + /* Since we set ATTR_*_SET flags above, and already done permission + * check, So don't let inode_change_ok() check it again. */ + 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; /* Extract epoch data if obtained. */ op_data->op_handle = md.body->handle; @@ -1219,13 +1271,15 @@ static int ll_setattr_done_writing(struct inode *inode, CDEBUG(D_INODE, "Epoch "LPU64" closed on "DFID" for truncate\n", op_data->op_ioepoch, PFID(&lli->lli_fid)); - op_data->op_flags = MF_EPOCH_CLOSE | MF_SOM_CHANGE; + op_data->op_flags = MF_EPOCH_CLOSE; + ll_done_writing_attr(inode, op_data); + ll_pack_inode2opdata(inode, op_data, NULL); + rc = md_done_writing(ll_i2sbi(inode)->ll_md_exp, op_data, mod); if (rc == -EAGAIN) { /* MDS has instructed us to obtain Size-on-MDS attribute * from OSTs and send setattr to back to MDS. */ - rc = ll_sizeonmds_update(inode, &op_data->op_handle, - op_data->op_ioepoch); + rc = ll_som_update(inode, op_data); } else if (rc) { CERROR("inode %lu mdc truncate failed: rc = %d\n", inode->i_ino, rc); @@ -1233,23 +1287,22 @@ static int ll_setattr_done_writing(struct inode *inode, RETURN(rc); } -static int ll_setattr_do_truncate(struct inode *inode, loff_t size) +static int ll_setattr_ost(struct inode *inode, struct iattr *attr) { - struct obd_capa *capa = ll_osscapa_get(inode, CAPA_OPC_OSS_TRUNC); + struct obd_capa *capa; int rc; - rc = cl_setattr_do_truncate(inode, size, capa); - ll_truncate_free_capa(capa); - return rc; -} + if (attr->ia_valid & ATTR_SIZE) + capa = ll_osscapa_get(inode, CAPA_OPC_OSS_TRUNC); + else + capa = ll_mdscapa_get(inode); -static int ll_setattr_ost(struct inode *inode) -{ - struct obd_capa *capa = ll_mdscapa_get(inode); - int rc; + rc = cl_setattr_ost(inode, attr, capa); - rc = cl_setattr_ost(inode, capa); - capa_put(capa); + if (attr->ia_valid & ATTR_SIZE) + ll_truncate_free_capa(capa); + else + capa_put(capa); return rc; } @@ -1267,8 +1320,9 @@ static int ll_setattr_ost(struct inode *inode) * 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; @@ -1279,12 +1333,20 @@ int ll_setattr_raw(struct inode *inode, struct iattr *attr) CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu valid %x\n", inode->i_ino, attr->ia_valid); - ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_SETATTR, 1); if (ia_valid & ATTR_SIZE) { + /* Check new size against VFS/VM file size limit and rlimit */ + rc = inode_newsize_ok(inode, attr->ia_size); + if (rc) + RETURN(rc); + + /* The maximum Lustre file size is variable, based on the + * OST maximum object size and number of stripes. This + * needs another check in addition to the VFS check above. */ if (attr->ia_size > ll_file_maxbytes(inode)) { - CDEBUG(D_INODE, "file too large %llu > "LPU64"\n", - attr->ia_size, ll_file_maxbytes(inode)); + CDEBUG(D_INODE,"file "DFID" too large %llu > "LPU64"\n", + PFID(&lli->lli_fid), attr->ia_size, + ll_file_maxbytes(inode)); RETURN(-EFBIG); } @@ -1292,35 +1354,25 @@ int ll_setattr_raw(struct inode *inode, struct iattr *attr) } /* POSIX: check before ATTR_*TIME_SET set (from inode_change_ok) */ - if (ia_valid & (ATTR_MTIME_SET | ATTR_ATIME_SET)) { - if (current->fsuid != inode->i_uid && + if (ia_valid & TIMES_SET_FLAGS) { + if (cfs_curproc_fsuid() != inode->i_uid && !cfs_capable(CFS_CAP_FOWNER)) RETURN(-EPERM); } /* We mark all of the fields "set" so MDS/OST does not re-set them */ if (attr->ia_valid & ATTR_CTIME) { - attr->ia_ctime = CURRENT_TIME; + attr->ia_ctime = CFS_CURRENT_TIME; attr->ia_valid |= ATTR_CTIME_SET; } if (!(ia_valid & ATTR_ATIME_SET) && (attr->ia_valid & ATTR_ATIME)) { - attr->ia_atime = CURRENT_TIME; + attr->ia_atime = CFS_CURRENT_TIME; attr->ia_valid |= ATTR_ATIME_SET; } if (!(ia_valid & ATTR_MTIME_SET) && (attr->ia_valid & ATTR_MTIME)) { - attr->ia_mtime = CURRENT_TIME; + attr->ia_mtime = CFS_CURRENT_TIME; attr->ia_valid |= ATTR_MTIME_SET; } - if ((attr->ia_valid & ATTR_CTIME) && !(attr->ia_valid & ATTR_MTIME)) { - /* To avoid stale mtime on mds, obtain it from ost and send - to mds. */ - rc = cl_glimpse_size(inode); - if (rc) - RETURN(rc); - - attr->ia_valid |= ATTR_MTIME_SET | ATTR_MTIME; - attr->ia_mtime = inode->i_mtime; - } if (attr->ia_valid & (ATTR_MTIME | ATTR_CTIME)) CDEBUG(D_INODE, "setting mtime %lu, ctime %lu, now = %lu\n", @@ -1339,14 +1391,24 @@ int ll_setattr_raw(struct inode *inode, 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); + } + memcpy(&op_data->op_attr, attr, sizeof(*attr)); /* Open epoch for truncate. */ - if ((ll_i2mdexp(inode)->exp_connect_flags & OBD_CONNECT_SOM) && - (ia_valid & ATTR_SIZE)) + if (exp_connect_som(ll_i2mdexp(inode)) && + (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); @@ -1357,39 +1419,62 @@ int ll_setattr_raw(struct inode *inode, struct iattr *attr) } if (ia_valid & ATTR_SIZE) - rc = ll_setattr_do_truncate(inode, attr->ia_size); - else if (ia_valid & (ATTR_MTIME | ATTR_MTIME_SET)) { - CDEBUG(D_INODE, "set mtime on OST inode %lu to %lu\n", - inode->i_ino, LTIME_S(attr->ia_mtime)); - rc = ll_setattr_ost(inode); - } + attr->ia_valid |= ATTR_SIZE; + if (ia_valid & (ATTR_SIZE | + ATTR_ATIME | ATTR_ATIME_SET | + ATTR_MTIME | ATTR_MTIME_SET)) + /* For truncate and utimes sending attributes to OSTs, setting + * mtime/atime to the 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 between MDT inode and OST objects */ + rc = ll_setattr_ost(inode, attr); EXIT; out: if (op_data) { - if (op_data->op_ioepoch) + if (op_data->op_ioepoch) { rc1 = ll_setattr_done_writing(inode, op_data, mod); + if (!rc) + rc = rc1; + } ll_finish_md_op_data(op_data); } - return rc ? rc : rc1; + if (!S_ISDIR(inode->i_mode)) + cfs_up_write(&lli->lli_trunc_sem); + + ll_stats_ops_tally(ll_i2sbi(inode), (ia_valid & ATTR_SIZE) ? + LPROC_LL_TRUNC : LPROC_LL_SETATTR, 1); + + return rc; } 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; - 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, @@ -1456,10 +1541,10 @@ 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; @@ -1491,9 +1576,11 @@ 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); - down(&lli->lli_size_sem); + cfs_down(&lli->lli_size_sem); LASSERT(lli->lli_size_sem_owner == NULL); lli->lli_size_sem_owner = current; lsm = lli->lli_smd; @@ -1516,24 +1603,7 @@ void ll_inode_size_unlock(struct inode *inode, int unlock_lsm) lov_stripe_unlock(lsm); LASSERT(lli->lli_size_sem_owner == current); lli->lli_size_sem_owner = NULL; - up(&lli->lli_size_sem); -} - -static void ll_replace_lsm(struct inode *inode, struct lov_stripe_md *lsm) -{ - struct ll_inode_info *lli = ll_i2info(inode); - - dump_lsm(D_INODE, lsm); - dump_lsm(D_INODE, lli->lli_smd); - LASSERTF(lsm->lsm_magic == LOV_MAGIC_JOIN, - "lsm must be joined lsm %p\n", lsm); - obd_free_memmd(ll_i2dtexp(inode), &lli->lli_smd); - CDEBUG(D_INODE, "replace lsm %p to lli_smd %p for inode %lu%u(%p)\n", - lsm, lli->lli_smd, inode->i_ino, inode->i_generation, inode); - lli->lli_smd = lsm; - lli->lli_maxbytes = lsm->lsm_maxbytes; - if (lli->lli_maxbytes > PAGE_CACHE_MAXBYTES) - lli->lli_maxbytes = PAGE_CACHE_MAXBYTES; + cfs_up(&lli->lli_size_sem); } void ll_update_inode(struct inode *inode, struct lustre_md *md) @@ -1545,40 +1615,42 @@ 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_mutex_lock(&lli->lli_och_mutex); if (lli->lli_smd == NULL) { if (lsm->lsm_magic != LOV_MAGIC_V1 && - lsm->lsm_magic != LOV_MAGIC_V3 && - lsm->lsm_magic != LOV_MAGIC_JOIN) { + 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_inode_init(inode, md); - /* ll_inode_size_lock() requires it is only - * called with lli_smd != NULL or lock_lsm == 0 - * or we can race between lock/unlock. - * bug 9547 */ + /* 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 > PAGE_CACHE_MAXBYTES) - lli->lli_maxbytes = PAGE_CACHE_MAXBYTES; + if (lli->lli_maxbytes > MAX_LFS_FILESIZE) + lli->lli_maxbytes = MAX_LFS_FILESIZE; } else { - if (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(); - } - } else { - cl_inode_init(inode, md); - ll_replace_lsm(inode, lsm); + 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) @@ -1591,33 +1663,34 @@ void ll_update_inode(struct inode *inode, struct lustre_md *md) } #ifdef CONFIG_FS_POSIX_ACL else if (body->valid & OBD_MD_FLACL) { - spin_lock(&lli->lli_lock); + cfs_spin_lock(&lli->lli_lock); if (lli->lli_posix_acl) posix_acl_release(lli->lli_posix_acl); lli->lli_posix_acl = md->posix_acl; - spin_unlock(&lli->lli_lock); + cfs_spin_unlock(&lli->lli_lock); } #endif - inode->i_ino = ll_fid_build_ino(sbi, &body->fid1); - inode->i_generation = ll_fid_build_gen(sbi, &body->fid1); - - if (body->valid & OBD_MD_FLATIME && - body->atime > LTIME_S(inode->i_atime)) - LTIME_S(inode->i_atime) = body->atime; - - /* mtime is always updated with ctime, but can be set in past. - As write and utime(2) may happen within 1 second, and utime's - mtime has a priority over write's one, so take mtime from mds - for the same ctimes. */ - if (body->valid & OBD_MD_FLCTIME && - body->ctime >= LTIME_S(inode->i_ctime)) { - LTIME_S(inode->i_ctime) = body->ctime; - if (body->valid & OBD_MD_FLMTIME) { - CDEBUG(D_INODE, "setting ino %lu mtime " - "from %lu to "LPU64"\n", inode->i_ino, + 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) { + if (body->atime > LTIME_S(inode->i_atime)) + LTIME_S(inode->i_atime) = body->atime; + lli->lli_lvb.lvb_atime = body->atime; + } + if (body->valid & OBD_MD_FLMTIME) { + if (body->mtime > LTIME_S(inode->i_mtime)) { + CDEBUG(D_INODE, "setting ino %lu mtime from %lu " + "to "LPU64"\n", inode->i_ino, LTIME_S(inode->i_mtime), body->mtime); LTIME_S(inode->i_mtime) = body->mtime; } + lli->lli_lvb.lvb_mtime = body->mtime; + } + if (body->valid & OBD_MD_FLCTIME) { + if (body->ctime > LTIME_S(inode->i_ctime)) + LTIME_S(inode->i_ctime) = body->ctime; + lli->lli_lvb.lvb_ctime = body->ctime; } if (body->valid & OBD_MD_FLMODE) inode->i_mode = (inode->i_mode & S_IFMT)|(body->mode & ~S_IFMT); @@ -1673,14 +1746,14 @@ 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); @@ -1688,7 +1761,10 @@ 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); } if (body->valid & OBD_MD_FLBLOCKS) @@ -1705,25 +1781,14 @@ 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; 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); @@ -1739,16 +1804,20 @@ 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; @@ -1759,9 +1828,6 @@ 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; } } @@ -1773,11 +1839,29 @@ 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; } @@ -1791,14 +1875,19 @@ int ll_iocontrol(struct inode *inode, struct file *file, ENTRY; switch(cmd) { - case EXT3_IOC_GETFLAGS: { + case FSFILT_IOC_GETFLAGS: { struct mdt_body *body; - struct obd_capa *oc; + struct md_op_data *op_data; - oc = ll_mdscapa_get(inode); - rc = md_getattr(sbi->ll_md_exp, ll_inode2fid(inode), oc, - OBD_MD_FLFLAGS, 0, &req); - capa_put(oc); + op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL, + 0, 0, LUSTRE_OPC_ANY, + NULL); + 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); + ll_finish_md_op_data(op_data); if (rc) { CERROR("failure %d inode %lu\n", rc, inode->i_ino); RETURN(-abs(rc)); @@ -1812,7 +1901,7 @@ int ll_iocontrol(struct inode *inode, struct file *file, RETURN(put_user(flags, (int *)arg)); } - case EXT3_IOC_SETFLAGS: { + case FSFILT_IOC_SETFLAGS: { struct lov_stripe_md *lsm = ll_i2info(inode)->lli_smd; struct obd_info oinfo = { { { 0 } } }; struct md_op_data *op_data; @@ -1836,32 +1925,36 @@ 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 || lsm == NULL) { + if (rc) { OBDO_FREE(oinfo.oi_oa); RETURN(rc); } + if (lsm == NULL) { + OBDO_FREE(oinfo.oi_oa); + GOTO(update_cache, rc); + } + 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); + 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("md_setattr_async fails: rc = %d\n", rc); + CERROR("osc_setattr_async fails: rc = %d\n",rc); RETURN(rc); } - inode->i_flags = ll_ext_to_inode_flags(flags | - MDS_BFLAG_EXT_FLAGS); - RETURN(0); + EXIT; +update_cache: + inode->i_flags = ll_ext_to_inode_flags(flags); + return 0; } default: RETURN(-ENOSYS); @@ -1874,7 +1967,7 @@ int ll_flush_ctx(struct inode *inode) { struct ll_sb_info *sbi = ll_i2sbi(inode); - CDEBUG(D_SEC, "flush context for user %d\n", current->uid); + CDEBUG(D_SEC, "flush context for user %d\n", cfs_curproc_uid()); obd_set_info_async(sbi->ll_md_exp, sizeof(KEY_FLUSH_CTX), KEY_FLUSH_CTX, @@ -1897,7 +1990,7 @@ void ll_umount_begin(struct super_block *sb) 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 = { 0 }; + struct obd_ioctl_data *ioc_data; ENTRY; #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT @@ -1921,8 +2014,6 @@ void ll_umount_begin(struct super_block *sb) return; } obd->obd_force = 1; - obd_iocontrol(IOC_OSC_SET_ACTIVE, sbi->ll_md_exp, sizeof ioc_data, - &ioc_data, NULL); obd = class_exp2obd(sbi->ll_dt_exp); if (obd == NULL) { @@ -1931,20 +2022,29 @@ void ll_umount_begin(struct super_block *sb) EXIT; return; } - obd->obd_force = 1; - obd_iocontrol(IOC_OSC_SET_ACTIVE, sbi->ll_dt_exp, sizeof ioc_data, - &ioc_data, NULL); + + OBD_ALLOC_PTR(ioc_data); + if (ioc_data) { + obd_iocontrol(IOC_OSC_SET_ACTIVE, sbi->ll_md_exp, + sizeof ioc_data, ioc_data, NULL); + + obd_iocontrol(IOC_OSC_SET_ACTIVE, sbi->ll_dt_exp, + sizeof ioc_data, ioc_data, NULL); + + OBD_FREE_PTR(ioc_data); + } + /* Really, we'd like to wait until there are no requests outstanding, * and then continue. For now, we just invalidate the requests, * schedule() and sleep one second if needed, and hope. */ - schedule(); + cfs_schedule(); #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT if (atomic_read(&vfsmnt->mnt_count) > 2) { - cfs_schedule_timeout(CFS_TASK_INTERRUPTIBLE, - cfs_time_seconds(1)); + cfs_schedule_timeout_and_set_state(CFS_TASK_INTERRUPTIBLE, + cfs_time_seconds(1)); if (atomic_read(&vfsmnt->mnt_count) > 2) LCONSOLE_WARN("Mount still busy with %d refs! You " "may try to umount it a bit later\n", @@ -1958,6 +2058,7 @@ 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; @@ -1968,8 +2069,9 @@ int ll_remount_fs(struct super_block *sb, int *flags, char *data) 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; } @@ -1977,6 +2079,10 @@ 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; } @@ -1992,9 +2098,6 @@ int ll_prep_inode(struct inode **inode, LASSERT(*inode || sb); sbi = sb ? ll_s2sbi(sb) : ll_i2sbi(*inode); - prune_deathrow(sbi, 1); - 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) @@ -2011,7 +2114,7 @@ int ll_prep_inode(struct inode **inode, */ LASSERT(fid_is_sane(&md.body->fid1)); - *inode = ll_iget(sb, ll_fid_build_ino(sbi, &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); @@ -2028,8 +2131,6 @@ int ll_prep_inode(struct inode **inode, } } - rc = obd_checkmd(sbi->ll_dt_exp, sbi->ll_md_exp, - ll_i2info(*inode)->lli_smd); out: md_free_lustre_md(sbi->ll_md_exp, &md); RETURN(rc); @@ -2056,8 +2157,14 @@ int ll_obd_statfs(struct inode *inode, void *arg) !data->ioc_pbuf1 || !data->ioc_pbuf2) GOTO(out_statfs, rc = -EINVAL); + if (data->ioc_inllen1 != sizeof(__u32) || + data->ioc_inllen2 != sizeof(__u32) || + data->ioc_plen1 != sizeof(struct obd_statfs) || + data->ioc_plen2 != sizeof(struct obd_uuid)) + 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; @@ -2135,14 +2242,29 @@ struct md_op_data * ll_prep_md_op_data(struct md_op_data *op_data, op_data->op_namelen = namelen; op_data->op_mode = mode; op_data->op_mod_time = cfs_time_current_sec(); - op_data->op_fsuid = current->fsuid; - op_data->op_fsgid = current->fsgid; + op_data->op_fsuid = cfs_curproc_fsuid(); + op_data->op_fsgid = cfs_curproc_fsgid(); op_data->op_cap = cfs_curproc_cap_pack(); op_data->op_bias = MDS_CHECK_SPLIT; op_data->op_opc = opc; 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; } @@ -2172,11 +2294,34 @@ 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"); 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); +}