X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Fllite%2Fdir.c;h=2549009e055cdec04f9323166c302297d183d41e;hb=498ee35323fcacf829f4e77c91e7700cb3660111;hp=456f6f1091c9aba57f69565556270a8fb466708c;hpb=e1d6fea43563624277b5cc1c91c084a6b52db7fe;p=fs%2Flustre-release.git diff --git a/lustre/llite/dir.c b/lustre/llite/dir.c index 456f6f1..2549009 100644 --- a/lustre/llite/dir.c +++ b/lustre/llite/dir.c @@ -1,731 +1,1266 @@ /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- * vim:expandtab:shiftwidth=8:tabstop=8: * - * Copyright (C) 1992, 1993, 1994, 1995 - * Remy Card (card@masi.ibp.fr) - * Laboratoire MASI - Institut Blaise Pascal - * Universite Pierre et Marie Curie (Paris VI) + * GPL HEADER START * - * from + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * - * linux/fs/minix/dir.c - * linux/fs/ext2/dir.c + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. * - * Copyright (C) 1991, 1992 Linus Torvalds + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). * - * ext2 directory handling functions + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf * - * Big-endian to little-endian byte-swapping/bitmaps by - * David S. Miller (davem@caip.rutgers.edu), 1995 + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. * - * All code that works with directory layout had been switched to pagecache - * and moved here. AV + * GPL HEADER END + */ +/* + * Copyright 2008 Sun Microsystems, Inc. All rights reserved + * Use is subject to license terms. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. * - * Adapted for Lustre Light - * Copyright (C) 2002, Cluster File Systems, Inc. + * lustre/llite/dir.c * + * Directory code for lustre client. */ #include -#include #include #include -#include +#include +#include #include +#include // for wait_on_buffer #define DEBUG_SUBSYSTEM S_LLITE -#include -#include -#include -#include -#include -#include -#include +#include +#include +#include +#include +#include +#include +#include +#include "llite_internal.h" + +#ifndef HAVE_PAGE_CHECKED +#ifdef HAVE_PG_FS_MISC +#define PageChecked(page) test_bit(PG_fs_misc, &(page)->flags) +#define SetPageChecked(page) set_bit(PG_fs_misc, &(page)->flags) +#else +#error PageChecked or PageFsMisc not defined in kernel +#endif +#endif -typedef struct ext2_dir_entry_2 ext2_dirent; - -#define PageChecked(page) test_bit(PG_checked, &(page)->flags) -#define SetPageChecked(page) set_bit(PG_checked, &(page)->flags) - - -static int ll_dir_prepare_write(struct file *file, struct page *page, unsigned from, unsigned to) -{ - return 0; -} +/* + * (new) readdir implementation overview. + * + * Original lustre readdir implementation cached exact copy of raw directory + * pages on the client. These pages were indexed in client page cache by + * logical offset in the directory file. This design, while very simple and + * intuitive had some inherent problems: + * + * . it implies that byte offset to the directory entry serves as a + * telldir(3)/seekdir(3) cookie, but that offset is not stable: in + * ext3/htree directory entries may move due to splits, and more + * importantly, + * + * . it is incompatible with the design of split directories for cmd3, + * that assumes that names are distributed across nodes based on their + * hash, and so readdir should be done in hash order. + * + * New readdir implementation does readdir in hash order, and uses hash of a + * file name as a telldir/seekdir cookie. This led to number of complications: + * + * . hash is not unique, so it cannot be used to index cached directory + * pages on the client (note, that it requires a whole pageful of hash + * collided entries to cause two pages to have identical hashes); + * + * . hash is not unique, so it cannot, strictly speaking, be used as an + * entry cookie. ext3/htree has the same problem and lustre implementation + * mimics their solution: seekdir(hash) positions directory at the first + * entry with the given hash. + * + * Client side. + * + * 0. caching + * + * Client caches directory pages using hash of the first entry as an index. As + * noted above hash is not unique, so this solution doesn't work as is: + * special processing is needed for "page hash chains" (i.e., sequences of + * pages filled with entries all having the same hash value). + * + * First, such chains have to be detected. To this end, server returns to the + * client the hash of the first entry on the page next to one returned. When + * client detects that this hash is the same as hash of the first entry on the + * returned page, page hash collision has to be handled. Pages in the + * hash chain, except first one, are termed "overflow pages". + * + * Solution to index uniqueness problem is to not cache overflow + * pages. Instead, when page hash collision is detected, all overflow pages + * from emerging chain are immediately requested from the server and placed in + * a special data structure (struct ll_dir_chain). This data structure is used + * by ll_readdir() to process entries from overflow pages. When readdir + * invocation finishes, overflow pages are discarded. If page hash collision + * chain weren't completely processed, next call to readdir will again detect + * page hash collision, again read overflow pages in, process next portion of + * entries and again discard the pages. This is not as wasteful as it looks, + * because, given reasonable hash, page hash collisions are extremely rare. + * + * 1. directory positioning + * + * When seekdir(hash) is called, original + * + * + * + * + * + * + * + * + * Server. + * + * identification of and access to overflow pages + * + * page format + * + * + * + * + * + */ /* returns the page unlocked, but with a reference */ static int ll_dir_readpage(struct file *file, struct page *page) { struct inode *inode = page->mapping->host; - struct ll_sb_info *sbi = ll_i2sbi(inode); - char *buf; - __u64 offset; - int rc = 0; struct ptlrpc_request *request; - struct lustre_handle lockh; - struct mds_body *body; - struct lookup_intent it = {IT_READDIR}; - + struct mdt_body *body; + struct obd_capa *oc; + __u64 hash; + int rc; ENTRY; - if ((inode->i_size + PAGE_CACHE_SIZE - 1) >> PAGE_SHIFT <= page->index){ - memset(kmap(page), 0, PAGE_CACHE_SIZE); - kunmap(page); - GOTO(readpage_out, rc); - } - - rc = ll_lock(inode, NULL, &it, &lockh); - request = (struct ptlrpc_request *)it.it_data; - ptlrpc_free_req(request); - if (rc != ELDLM_OK) { - CERROR("lock enqueue: err: %d\n", rc); - UnlockPage(page); - RETURN(rc); - } - ldlm_lock_dump((void *)(unsigned long)lockh.addr); - - if (Page_Uptodate(page)) { - CERROR("Explain this please?\n"); - GOTO(readpage_out, rc); + hash = (__u64)hash_x_index(page->index); + CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p) off %lu\n", + inode->i_ino, inode->i_generation, inode, (unsigned long)hash); + + oc = ll_mdscapa_get(inode); + rc = md_readpage(ll_i2sbi(inode)->ll_md_exp, ll_inode2fid(inode), + oc, hash, page, &request); + capa_put(oc); + if (!rc) { + body = req_capsule_server_get(&request->rq_pill, &RMF_MDT_BODY); + /* Checked by mdc_readpage() */ + LASSERT(body != NULL); + + if (body->valid & OBD_MD_FLSIZE) { + ll_inode_size_lock(inode, 0); + i_size_write(inode, body->size); + ll_inode_size_unlock(inode, 0); + } + SetPageUptodate(page); } + ptlrpc_req_finished(request); - offset = page->index << PAGE_SHIFT; - buf = kmap(page); - rc = mdc_readpage(&sbi->ll_mdc_conn, inode->i_ino, - S_IFDIR, offset, buf, &request); - kunmap(page); - body = lustre_msg_buf(request->rq_repmsg, 0); - if (!body) - rc = -EINVAL; - if (body) - inode->i_size = body->size; - ptlrpc_free_req(request); + unlock_page(page); EXIT; - - readpage_out: - if ( !rc ) - SetPageUptodate(page); - - UnlockPage(page); - rc = ll_unlock(LCK_PR, &lockh); - if (rc != ELDLM_OK) - CERROR("ll_unlock: err: %d\n", rc); return rc; -} /* ll_dir_readpage */ +} struct address_space_operations ll_dir_aops = { - readpage: ll_dir_readpage, - prepare_write: ll_dir_prepare_write + .readpage = ll_dir_readpage, }; -int waitfor_one_page(struct page *page) +static void ll_check_page(struct inode *dir, struct page *page) { - int error = 0; - struct buffer_head *bh, *head = page->buffers; - - bh = head; - do { - wait_on_buffer(bh); - if (buffer_req(bh) && !buffer_uptodate(bh)) - error = -EIO; - } while ((bh = bh->b_this_page) != head); - return error; -} - -/* - * ext2 uses block-sized chunks. Arguably, sector-sized ones would be - * more robust, but we have what we have - */ -static inline unsigned ext2_chunk_size(struct inode *inode) -{ - return inode->i_sb->s_blocksize; + /* XXX: check page format later */ + SetPageChecked(page); } -static inline void ext2_put_page(struct page *page) +static void ll_release_page(struct page *page, __u64 hash, + __u64 start, __u64 end) { kunmap(page); + lock_page(page); + if (likely(page->mapping != NULL)) { + ll_truncate_complete_page(page); + unlock_page(page); + } else { + unlock_page(page); + CWARN("NULL mapping page %p, truncated by others: " + "hash("LPX64") | start("LPX64") | end("LPX64")\n", + page, hash, start, end); + } page_cache_release(page); } -static inline unsigned long dir_pages(struct inode *inode) +/* + * Find, kmap and return page that contains given hash. + */ +static struct page *ll_dir_page_locate(struct inode *dir, __u64 hash, + __u64 *start, __u64 *end) { - return (inode->i_size+PAGE_CACHE_SIZE-1)>>PAGE_CACHE_SHIFT; -} - -extern void set_page_clean(struct page *page); + struct address_space *mapping = dir->i_mapping; + /* + * Complement of hash is used as an index so that + * radix_tree_gang_lookup() can be used to find a page with starting + * hash _smaller_ than one we are looking for. + */ + unsigned long offset = hash_x_index((unsigned long)hash); + struct page *page; + int found; + + TREE_READ_LOCK_IRQ(mapping); + found = radix_tree_gang_lookup(&mapping->page_tree, + (void **)&page, offset, 1); + if (found > 0) { + struct lu_dirpage *dp; + + page_cache_get(page); + TREE_READ_UNLOCK_IRQ(mapping); + /* + * In contrast to find_lock_page() we are sure that directory + * page cannot be truncated (while DLM lock is held) and, + * hence, can avoid restart. + * + * In fact, page cannot be locked here at all, because + * ll_dir_readpage() does synchronous io. + */ + wait_on_page(page); + if (PageUptodate(page)) { + dp = kmap(page); + *start = le64_to_cpu(dp->ldp_hash_start); + *end = le64_to_cpu(dp->ldp_hash_end); + LASSERT(*start <= hash); + if (hash > *end || (*end != *start && hash == *end)) { + ll_release_page(page, hash, *start, *end); + page = NULL; + } + } else { + page_cache_release(page); + page = ERR_PTR(-EIO); + } -static int ext2_commit_chunk(struct page *page, unsigned from, unsigned to) -{ - struct inode *dir = page->mapping->host; - loff_t new_size = (page->index << PAGE_CACHE_SHIFT) + to; - int err = 0; - - dir->i_version = ++event; - if (new_size > dir->i_size) - dir->i_size = new_size; - SetPageUptodate(page); - set_page_clean(page); - - //page->mapping->a_ops->commit_write(NULL, page, from, to); - //if (IS_SYNC(dir)) - // err = waitfor_one_page(page); - return err; + } else { + TREE_READ_UNLOCK_IRQ(mapping); + page = NULL; + } + return page; } -static void ext2_check_page(struct page *page) +struct page *ll_get_dir_page(struct inode *dir, __u64 hash, int exact, + struct ll_dir_chain *chain) { - struct inode *dir = page->mapping->host; - unsigned chunk_size = ext2_chunk_size(dir); - char *kaddr = page_address(page); - // u32 max_inumber = le32_to_cpu(sb->u.ext2_sb.s_es->s_inodes_count); - unsigned offs, rec_len; - unsigned limit = PAGE_CACHE_SIZE; - ext2_dirent *p; - char *error; - - if ((dir->i_size >> PAGE_CACHE_SHIFT) == page->index) { - limit = dir->i_size & ~PAGE_CACHE_MASK; - if (limit & (chunk_size - 1)) { - CERROR("limit %d dir size %lld index %ld\n", - limit, dir->i_size, page->index); - goto Ebadsize; + ldlm_policy_data_t policy = {.l_inodebits = {MDS_INODELOCK_UPDATE} }; + struct address_space *mapping = dir->i_mapping; + struct lustre_handle lockh; + struct lu_dirpage *dp; + struct page *page; + ldlm_mode_t mode; + int rc; + __u64 start = 0; + __u64 end = 0; + + mode = LCK_PR; + rc = md_lock_match(ll_i2sbi(dir)->ll_md_exp, LDLM_FL_BLOCK_GRANTED, + ll_inode2fid(dir), LDLM_IBITS, &policy, mode, &lockh); + if (!rc) { + struct ldlm_enqueue_info einfo = { LDLM_IBITS, mode, + ll_md_blocking_ast, ldlm_completion_ast, + NULL, NULL, dir }; + struct lookup_intent it = { .it_op = IT_READDIR }; + struct ptlrpc_request *request; + struct md_op_data *op_data; + + op_data = ll_prep_md_op_data(NULL, dir, NULL, NULL, 0, 0, + LUSTRE_OPC_ANY, NULL); + if (IS_ERR(op_data)) + return (void *)op_data; + + rc = md_enqueue(ll_i2sbi(dir)->ll_md_exp, &einfo, &it, + op_data, &lockh, NULL, 0, NULL, 0); + + ll_finish_md_op_data(op_data); + + request = (struct ptlrpc_request *)it.d.lustre.it_data; + if (request) + ptlrpc_req_finished(request); + if (rc < 0) { + CERROR("lock enqueue: rc: %d\n", rc); + return ERR_PTR(rc); } - for (offs = limit; offsrec_len = cpu_to_le16(chunk_size); - p->name_len = 0; - p->inode = 0; + } else { + /* for cross-ref object, l_ast_data of the lock may not be set, + * we reset it here */ + md_set_lock_data(ll_i2sbi(dir)->ll_md_exp, &lockh.cookie, + dir, NULL); + } + ldlm_lock_dump_handle(D_OTHER, &lockh); + + page = ll_dir_page_locate(dir, hash, &start, &end); + if (IS_ERR(page)) + GOTO(out_unlock, page); + + if (page != NULL) { + /* + * XXX nikita: not entirely correct handling of a corner case: + * suppose hash chain of entries with hash value HASH crosses + * border between pages P0 and P1. First both P0 and P1 are + * cached, seekdir() is called for some entry from the P0 part + * of the chain. Later P0 goes out of cache. telldir(HASH) + * happens and finds P1, as it starts with matching hash + * value. Remaining entries from P0 part of the chain are + * skipped. (Is that really a bug?) + * + * Possible solutions: 0. don't cache P1 is such case, handle + * it as an "overflow" page. 1. invalidate all pages at + * once. 2. use HASH|1 as an index for P1. + */ + if (exact && hash != start) { + /* + * readdir asked for a page starting _exactly_ from + * given hash, but cache contains stale page, with + * entries with smaller hash values. Stale page should + * be invalidated, and new one fetched. + */ + CWARN("Stale readpage page %p: "LPX64" != "LPX64"\n", + page, hash, start); + ll_release_page(page, hash, start, end); + } else { + GOTO(hash_collision, page); } - if (!limit) - goto out; } - for (offs = 0; offs <= limit - EXT2_DIR_REC_LEN(1); offs += rec_len) { - p = (ext2_dirent *)(kaddr + offs); - rec_len = le16_to_cpu(p->rec_len); - - if (rec_len < EXT2_DIR_REC_LEN(1)) - goto Eshort; - if (rec_len & 3) - goto Ealign; - if (rec_len < EXT2_DIR_REC_LEN(p->name_len)) - goto Enamelen; - if (((offs + rec_len - 1) ^ offs) & ~(chunk_size-1)) - goto Espan; - // if (le32_to_cpu(p->inode) > max_inumber) - //goto Einumber; + + page = read_cache_page(mapping, hash_x_index((unsigned long)hash), + (filler_t*)mapping->a_ops->readpage, NULL); + if (IS_ERR(page)) + GOTO(out_unlock, page); + + wait_on_page(page); + (void)kmap(page); + if (!PageUptodate(page)) + goto fail; + if (!PageChecked(page)) + ll_check_page(dir, page); + if (PageError(page)) + goto fail; +hash_collision: + dp = page_address(page); + + start = le64_to_cpu(dp->ldp_hash_start); + end = le64_to_cpu(dp->ldp_hash_end); + if (end == start) { + LASSERT(start == hash); + CWARN("Page-wide hash collision: %#lx\n", (unsigned long)end); + /* + * Fetch whole overflow chain... + * + * XXX not yet. + */ + goto fail; } - if (offs != limit) - goto Eend; -out: - SetPageChecked(page); - return; - - /* Too bad, we had an error */ - -Ebadsize: - CERROR("ext2_check_page" - "size of directory #%lu is not a multiple of chunk size\n", - dir->i_ino - ); - goto fail; -Eshort: - error = "rec_len is smaller than minimal"; - goto bad_entry; -Ealign: - error = "unaligned directory entry"; - goto bad_entry; -Enamelen: - error = "rec_len is too small for name_len"; - goto bad_entry; -Espan: - error = "directory entry across blocks"; - goto bad_entry; - //Einumber: - // error = "inode out of bounds"; -bad_entry: - CERROR("ext2_check_page: bad entry in directory #%lu: %s - " - "offset=%lu, inode=%lu, rec_len=%d, name_len=%d", - dir->i_ino, error, (page->index<inode), - rec_len, p->name_len); - goto fail; -Eend: - p = (ext2_dirent *)(kaddr + offs); - CERROR("ext2_check_page" - "entry in directory #%lu spans the page boundary" - "offset=%lu, inode=%lu", - dir->i_ino, (page->index<inode)); +out_unlock: + ldlm_lock_decref(&lockh, mode); + return page; + fail: - SetPageChecked(page); - SetPageError(page); - LBUG(); + ll_put_page(page); + page = ERR_PTR(-EIO); + goto out_unlock; } -static struct page * ext2_get_page(struct inode *dir, unsigned long n) +int ll_readdir(struct file *filp, void *cookie, filldir_t filldir) { - struct address_space *mapping = dir->i_mapping; - struct page *page = read_cache_page(mapping, n, - (filler_t*)mapping->a_ops->readpage, NULL); - if (!IS_ERR(page)) { - wait_on_page(page); - kmap(page); - if (!Page_Uptodate(page)) - goto fail; - if (!PageChecked(page)) - ext2_check_page(page); - if (PageError(page)) - goto fail; + struct inode *inode = filp->f_dentry->d_inode; + struct ll_inode_info *info = ll_i2info(inode); + struct ll_sb_info *sbi = ll_i2sbi(inode); + __u64 pos = filp->f_pos; + struct page *page; + struct ll_dir_chain chain; + int rc; + int done; + int shift; + __u16 type; + ENTRY; + + CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p) pos %lu/%llu\n", + inode->i_ino, inode->i_generation, inode, + (unsigned long)pos, i_size_read(inode)); + + if (pos == DIR_END_OFF) + /* + * end-of-file. + */ + RETURN(0); + + rc = 0; + done = 0; + shift = 0; + ll_dir_chain_init(&chain); + + page = ll_get_dir_page(inode, pos, 0, &chain); + + while (rc == 0 && !done) { + struct lu_dirpage *dp; + struct lu_dirent *ent; + + if (!IS_ERR(page)) { + /* + * If page is empty (end of directory is reached), + * use this value. + */ + __u64 hash = DIR_END_OFF; + __u64 next; + + dp = page_address(page); + for (ent = lu_dirent_start(dp); ent != NULL && !done; + ent = lu_dirent_next(ent)) { + char *name; + int namelen; + struct lu_fid fid; + ino_t ino; + + /* + * XXX: implement correct swabbing here. + */ + + hash = le64_to_cpu(ent->lde_hash); + namelen = le16_to_cpu(ent->lde_namelen); + + if (hash < pos) + /* + * Skip until we find target hash + * value. + */ + continue; + + if (namelen == 0) + /* + * Skip dummy record. + */ + continue; + + fid = ent->lde_fid; + name = ent->lde_name; + fid_le_to_cpu(&fid, &fid); + ino = ll_fid_build_ino(sbi, &fid); + type = ll_dirent_type_get(ent); + done = filldir(cookie, name, namelen, + (loff_t)hash, ino, type); + } + next = le64_to_cpu(dp->ldp_hash_end); + ll_put_page(page); + if (!done) { + pos = next; + if (pos == DIR_END_OFF) + /* + * End of directory reached. + */ + done = 1; + else if (1 /* chain is exhausted*/) + /* + * Normal case: continue to the next + * page. + */ + page = ll_get_dir_page(inode, pos, 1, + &chain); + else { + /* + * go into overflow page. + */ + } + } else + pos = hash; + } else { + rc = PTR_ERR(page); + CERROR("error reading dir "DFID" at %lu: rc %d\n", + PFID(&info->lli_fid), (unsigned long)pos, rc); + } } - return page; -fail: - ext2_put_page(page); - return ERR_PTR(-EIO); + filp->f_pos = (loff_t)pos; + filp->f_version = inode->i_version; + touch_atime(filp->f_vfsmnt, filp->f_dentry); + + ll_dir_chain_fini(&chain); + + RETURN(rc); } -/* - * NOTE! unlike strncmp, ext2_match returns 1 for success, 0 for failure. - * - * len <= EXT2_NAME_LEN and de != NULL are guaranteed by caller. - */ -static inline int ext2_match (int len, const char * const name, - struct ext2_dir_entry_2 * de) +int ll_send_mgc_param(struct obd_export *mgc, char *string) { - if (len != de->name_len) - return 0; - if (!de->inode) - return 0; - return !memcmp(name, de->name, len); + struct mgs_send_param *msp; + int rc = 0; + + OBD_ALLOC_PTR(msp); + if (!msp) + return -ENOMEM; + + strncpy(msp->mgs_param, string, MGS_PARAM_MAXLEN); + rc = obd_set_info_async(mgc, sizeof(KEY_SET_INFO), KEY_SET_INFO, + sizeof(struct mgs_send_param), msp, NULL); + if (rc) + CERROR("Failed to set parameter: %d\n", rc); + OBD_FREE_PTR(msp); + + return rc; } -/* - * p is at least 6 bytes before the end of page - */ -static inline ext2_dirent *ext2_next_entry(ext2_dirent *p) +char *ll_get_fsname(struct inode *inode) { - return (ext2_dirent *)((char*)p + le16_to_cpu(p->rec_len)); + struct lustre_sb_info *lsi = s2lsi(inode->i_sb); + char *ptr, *fsname; + int len; + + OBD_ALLOC(fsname, MGS_PARAM_MAXLEN); + len = strlen(lsi->lsi_lmd->lmd_profile); + ptr = strrchr(lsi->lsi_lmd->lmd_profile, '-'); + if (ptr && (strcmp(ptr, "-client") == 0)) + len -= 7; + strncpy(fsname, lsi->lsi_lmd->lmd_profile, len); + fsname[len] = '\0'; + + return fsname; } -static inline unsigned -ext2_validate_entry(char *base, unsigned offset, unsigned mask) +int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump, + int set_default) { - ext2_dirent *de = (ext2_dirent*)(base + offset); - ext2_dirent *p = (ext2_dirent*)(base + (offset&mask)); - while ((char*)p < (char*)de) - p = ext2_next_entry(p); - return (char *)p - base; + struct ll_sb_info *sbi = ll_i2sbi(inode); + struct md_op_data *op_data; + struct ptlrpc_request *req = NULL; + int rc = 0; + struct lustre_sb_info *lsi = s2lsi(inode->i_sb); + struct obd_device *mgc = lsi->lsi_mgc; + char *fsname = NULL, *param = NULL; + int lum_size; + + /* + * This is coming from userspace, so should be in + * local endian. But the MDS would like it in little + * endian, so we swab it before we send it. + */ + switch (lump->lmm_magic) { + case LOV_USER_MAGIC_V1: { + if (lump->lmm_magic != cpu_to_le32(LOV_USER_MAGIC_V1)) + lustre_swab_lov_user_md_v1(lump); + lum_size = sizeof(struct lov_user_md_v1); + break; + } + case LOV_USER_MAGIC_V3: { + if (lump->lmm_magic != cpu_to_le32(LOV_USER_MAGIC_V3)) + lustre_swab_lov_user_md_v3((struct lov_user_md_v3 *)lump); + lum_size = sizeof(struct lov_user_md_v3); + break; + } + default: { + CDEBUG(D_IOCTL, "bad userland LOV MAGIC:" + " %#08x != %#08x nor %#08x\n", + lump->lmm_magic, LOV_USER_MAGIC_V1, + LOV_USER_MAGIC_V3); + RETURN(-EINVAL); + } + } + + 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)); + + /* swabbing is done in lov_setstripe() on server side */ + rc = md_setattr(sbi->ll_md_exp, op_data, lump, lum_size, + NULL, 0, &req, NULL); + ll_finish_md_op_data(op_data); + ptlrpc_req_finished(req); + if (rc) { + if (rc != -EPERM && rc != -EACCES) + CERROR("mdc_setattr fails: rc = %d\n", rc); + } + + /* In the following we use the fact that LOV_USER_MAGIC_V1 and + LOV_USER_MAGIC_V3 have the same initial fields so we do not + need the make the distiction between the 2 versions */ + if (set_default && mgc->u.cli.cl_mgc_mgsexp) { + OBD_ALLOC(param, MGS_PARAM_MAXLEN); + + /* Get fsname and assume devname to be -MDT0000. */ + fsname = ll_get_fsname(inode); + /* Set root stripesize */ + sprintf(param, "%s-MDT0000.lov.stripesize=%u", fsname, + lump->lmm_stripe_size); + rc = ll_send_mgc_param(mgc->u.cli.cl_mgc_mgsexp, param); + if (rc) + goto end; + + /* Set root stripecount */ + sprintf(param, "%s-MDT0000.lov.stripecount=%hd", fsname, + lump->lmm_stripe_count); + rc = ll_send_mgc_param(mgc->u.cli.cl_mgc_mgsexp, param); + if (rc) + goto end; + + /* Set root stripeoffset */ + sprintf(param, "%s-MDT0000.lov.stripeoffset=%hd", fsname, + lump->lmm_stripe_offset); + rc = ll_send_mgc_param(mgc->u.cli.cl_mgc_mgsexp, param); + if (rc) + goto end; +end: + if (fsname) + OBD_FREE(fsname, MGS_PARAM_MAXLEN); + if (param) + OBD_FREE(param, MGS_PARAM_MAXLEN); + } + return rc; } -static unsigned char ext2_filetype_table[EXT2_FT_MAX] = { - [EXT2_FT_UNKNOWN] DT_UNKNOWN, - [EXT2_FT_REG_FILE] DT_REG, - [EXT2_FT_DIR] DT_DIR, - [EXT2_FT_CHRDEV] DT_CHR, - [EXT2_FT_BLKDEV] DT_BLK, - [EXT2_FT_FIFO] DT_FIFO, - [EXT2_FT_SOCK] DT_SOCK, - [EXT2_FT_SYMLINK] DT_LNK, -}; +int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmmp, + int *lmm_size, struct ptlrpc_request **request) +{ + struct ll_sb_info *sbi = ll_i2sbi(inode); + struct mdt_body *body; + struct lov_mds_md *lmm = NULL; + struct ptlrpc_request *req = NULL; + int rc, lmmsize; + struct obd_capa *oc; + + rc = ll_get_max_mdsize(sbi, &lmmsize); + if (rc) + RETURN(rc); -static unsigned int ll_dt2fmt[DT_WHT + 1] = { - [EXT2_FT_UNKNOWN] 0, - [EXT2_FT_REG_FILE] S_IFREG, - [EXT2_FT_DIR] S_IFDIR, - [EXT2_FT_CHRDEV] S_IFCHR, - [EXT2_FT_BLKDEV] S_IFBLK, - [EXT2_FT_FIFO] S_IFIFO, - [EXT2_FT_SOCK] S_IFSOCK, - [EXT2_FT_SYMLINK] S_IFLNK -}; + oc = ll_mdscapa_get(inode); + rc = md_getattr(sbi->ll_md_exp, ll_inode2fid(inode), + oc, OBD_MD_FLEASIZE | OBD_MD_FLDIREA, + lmmsize, &req); + capa_put(oc); + if (rc < 0) { + CDEBUG(D_INFO, "md_getattr failed on inode " + "%lu/%u: rc %d\n", inode->i_ino, + inode->i_generation, rc); + GOTO(out, rc); + } -#define S_SHIFT 12 -static unsigned char ext2_type_by_mode[S_IFMT >> S_SHIFT] = { - [S_IFREG >> S_SHIFT] EXT2_FT_REG_FILE, - [S_IFDIR >> S_SHIFT] EXT2_FT_DIR, - [S_IFCHR >> S_SHIFT] EXT2_FT_CHRDEV, - [S_IFBLK >> S_SHIFT] EXT2_FT_BLKDEV, - [S_IFIFO >> S_SHIFT] EXT2_FT_FIFO, - [S_IFSOCK >> S_SHIFT] EXT2_FT_SOCK, - [S_IFLNK >> S_SHIFT] EXT2_FT_SYMLINK, -}; + body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY); + LASSERT(body != NULL); -static inline void ext2_set_de_type(ext2_dirent *de, struct inode *inode) -{ - mode_t mode = inode->i_mode; - de->file_type = ext2_type_by_mode[(mode & S_IFMT)>>S_SHIFT]; + lmmsize = body->eadatasize; + + if (!(body->valid & (OBD_MD_FLEASIZE | OBD_MD_FLDIREA)) || + lmmsize == 0) { + GOTO(out, rc = -ENODATA); + } + + lmm = req_capsule_server_sized_get(&req->rq_pill, + &RMF_MDT_MD, lmmsize); + LASSERT(lmm != NULL); + + /* + * This is coming from the MDS, so is probably in + * little endian. We convert it to host endian before + * passing it to userspace. + */ + /* We don't swab objects for directories */ + switch (le32_to_cpu(lmm->lmm_magic)) { + case LOV_MAGIC_V1: + if (LOV_MAGIC != cpu_to_le32(LOV_MAGIC)) + lustre_swab_lov_user_md_v1((struct lov_user_md_v1 *)lmm); + break; + case LOV_MAGIC_V3: + if (LOV_MAGIC != cpu_to_le32(LOV_MAGIC)) + lustre_swab_lov_user_md_v3((struct lov_user_md_v3 *)lmm); + break; + default: + CERROR("unknown magic: %lX\n", (unsigned long)lmm->lmm_magic); + rc = -EPROTO; + } +out: + *lmmp = lmm; + *lmm_size = lmmsize; + *request = req; + return rc; } -int ll_readdir(struct file * filp, void * dirent, filldir_t filldir) +static int ll_dir_ioctl(struct inode *inode, struct file *file, + unsigned int cmd, unsigned long arg) { - loff_t pos = filp->f_pos; - struct inode *inode = filp->f_dentry->d_inode; - // XXX struct super_block *sb = inode->i_sb; - unsigned offset = pos & ~PAGE_CACHE_MASK; - unsigned long n = pos >> PAGE_CACHE_SHIFT; - unsigned long npages = dir_pages(inode); - unsigned chunk_mask = ~(ext2_chunk_size(inode)-1); - unsigned char *types = NULL; - int need_revalidate = (filp->f_version != inode->i_version); + struct ll_sb_info *sbi = ll_i2sbi(inode); + struct obd_ioctl_data *data; ENTRY; - if (pos > inode->i_size - EXT2_DIR_REC_LEN(1)) - GOTO(done, 0); - - types = ext2_filetype_table; + CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p), cmd=%#x\n", + inode->i_ino, inode->i_generation, inode, cmd); + + /* asm-ppc{,64} declares TCGETS, et. al. as type 't' not 'T' */ + if (_IOC_TYPE(cmd) == 'T' || _IOC_TYPE(cmd) == 't') /* tty ioctls */ + return -ENOTTY; + + ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_IOCTL, 1); + switch(cmd) { + case EXT3_IOC_GETFLAGS: + case EXT3_IOC_SETFLAGS: + RETURN(ll_iocontrol(inode, file, cmd, arg)); + case EXT3_IOC_GETVERSION_OLD: + case EXT3_IOC_GETVERSION: + RETURN(put_user(inode->i_generation, (int *)arg)); + /* We need to special case any other ioctls we want to handle, + * to send them to the MDS/OST as appropriate and to properly + * network encode the arg field. + case EXT3_IOC_SETVERSION_OLD: + case EXT3_IOC_SETVERSION: + */ + case IOC_MDC_LOOKUP: { + struct ptlrpc_request *request = NULL; + int namelen, rc, len = 0; + char *buf = NULL; + char *filename; + struct obd_capa *oc; + + rc = obd_ioctl_getdata(&buf, &len, (void *)arg); + if (rc) + RETURN(rc); + data = (void *)buf; + + filename = data->ioc_inlbuf1; + namelen = data->ioc_inllen1; + + if (namelen < 1) { + CDEBUG(D_INFO, "IOC_MDC_LOOKUP missing filename\n"); + GOTO(out, rc = -EINVAL); + } - for ( ; n < npages; n++, offset = 0) { - char *kaddr, *limit; - ext2_dirent *de; - struct page *page = ext2_get_page(inode, n); + oc = ll_mdscapa_get(inode); + rc = md_getattr_name(sbi->ll_md_exp, ll_inode2fid(inode), oc, + filename, namelen, OBD_MD_FLID, 0, + ll_i2suppgid(inode), &request); + capa_put(oc); + if (rc < 0) { + CDEBUG(D_INFO, "md_getattr_name: %d\n", rc); + GOTO(out, rc); + } - /* size might have been updated by mdc_readpage */ - npages = dir_pages(inode); + ptlrpc_req_finished(request); - if (IS_ERR(page)) - continue; - kaddr = page_address(page); - if (need_revalidate) { - offset = ext2_validate_entry(kaddr, offset, chunk_mask); - need_revalidate = 0; + EXIT; + out: + obd_ioctl_freedata(buf, len); + return rc; + } + case LL_IOC_LOV_SETSTRIPE: { + struct lov_user_md_v3 lumv3; + struct lov_user_md_v1 *lumv1 = (struct lov_user_md_v1 *)&lumv3; + struct lov_user_md_v1 *lumv1p = (struct lov_user_md_v1 *)arg; + struct lov_user_md_v3 *lumv3p = (struct lov_user_md_v3 *)arg; + + int rc = 0; + int set_default = 0; + + LASSERT(sizeof(lumv3) == sizeof(*lumv3p)); + LASSERT(sizeof(lumv3.lmm_objects[0]) == + sizeof(lumv3p->lmm_objects[0])); + /* first try with v1 which is smaller than v3 */ + if (copy_from_user(lumv1, lumv1p, sizeof(*lumv1))) + RETURN(-EFAULT); + + if (lumv1->lmm_magic == LOV_USER_MAGIC_V3) { + if (copy_from_user(&lumv3, lumv3p, sizeof(lumv3))) + RETURN(-EFAULT); } - de = (ext2_dirent *)(kaddr+offset); - limit = kaddr + PAGE_CACHE_SIZE - EXT2_DIR_REC_LEN(1); - for ( ;(char*)de <= limit; de = ext2_next_entry(de)) - if (de->inode) { - int over; - unsigned char d_type = DT_UNKNOWN; - - if (types && de->file_type < EXT2_FT_MAX) - d_type = types[de->file_type]; - - offset = (char *)de - kaddr; - over = filldir(dirent, de->name, de->name_len, - (n<inode), d_type); - if (over) { - ext2_put_page(page); - GOTO(done,0); - } - } - ext2_put_page(page); + + if (inode->i_sb->s_root == file->f_dentry) + set_default = 1; + + /* in v1 and v3 cases lumv1 points to data */ + rc = ll_dir_setstripe(inode, lumv1, set_default); + + RETURN(rc); } + case LL_IOC_OBD_STATFS: + RETURN(ll_obd_statfs(inode, (void *)arg)); + case LL_IOC_LOV_GETSTRIPE: + case LL_IOC_MDC_GETINFO: + case IOC_MDC_GETFILEINFO: + case IOC_MDC_GETFILESTRIPE: { + struct ptlrpc_request *request = NULL; + struct lov_user_md *lump; + struct lov_mds_md *lmm = NULL; + struct mdt_body *body; + char *filename = NULL; + int rc, lmmsize; + + if (cmd == IOC_MDC_GETFILEINFO || + cmd == IOC_MDC_GETFILESTRIPE) { + filename = getname((const char *)arg); + if (IS_ERR(filename)) + RETURN(PTR_ERR(filename)); + + rc = ll_lov_getstripe_ea_info(inode, filename, &lmm, + &lmmsize, &request); + } else { + rc = ll_dir_getstripe(inode, &lmm, &lmmsize, &request); + } -done: - filp->f_pos = (n << PAGE_CACHE_SHIFT) | offset; - filp->f_version = inode->i_version; - UPDATE_ATIME(inode); - RETURN(0); -} + if (request) { + body = req_capsule_server_get(&request->rq_pill, + &RMF_MDT_BODY); + LASSERT(body != NULL); + } else { + GOTO(out_req, rc); + } -/* - * ext2_find_entry() - * - * finds an entry in the specified directory with the wanted name. It - * returns the page in which the entry was found, and the entry itself - * (as a parameter - res_dir). Page is returned mapped and unlocked. - * Entry is guaranteed to be valid. - */ -struct ext2_dir_entry_2 * ext2_find_entry (struct inode * dir, - struct dentry *dentry, struct page ** res_page) -{ - const char *name = dentry->d_name.name; - int namelen = dentry->d_name.len; - unsigned reclen = EXT2_DIR_REC_LEN(namelen); - unsigned long start, n; - unsigned long npages = dir_pages(dir); - struct page *page = NULL; - ext2_dirent * de; - - /* OFFSET_CACHE */ - *res_page = NULL; - - // start = dir->u.ext2_i.i_dir_start_lookup; - start = 0; - if (start >= npages) - start = 0; - n = start; - do { - char *kaddr; - page = ext2_get_page(dir, n); - if (!IS_ERR(page)) { - kaddr = page_address(page); - de = (ext2_dirent *) kaddr; - kaddr += PAGE_CACHE_SIZE - reclen; - while ((char *) de <= kaddr) { - if (ext2_match (namelen, name, de)) - goto found; - de = ext2_next_entry(de); - } - ext2_put_page(page); + if (rc < 0) { + if (rc == -ENODATA && (cmd == IOC_MDC_GETFILEINFO || + cmd == LL_IOC_MDC_GETINFO)) + GOTO(skip_lmm, rc = 0); + else + GOTO(out_req, rc); } - if (++n >= npages) - n = 0; - } while (n != start); - return NULL; - -found: - *res_page = page; - // dir->u.ext2_i.i_dir_start_lookup = n; - return de; -} -struct ext2_dir_entry_2 * ext2_dotdot (struct inode *dir, struct page **p) -{ - struct page *page = ext2_get_page(dir, 0); - ext2_dirent *de = NULL; + if (cmd == IOC_MDC_GETFILESTRIPE || + cmd == LL_IOC_LOV_GETSTRIPE) { + lump = (struct lov_user_md *)arg; + } else { + struct lov_user_mds_data *lmdp; + lmdp = (struct lov_user_mds_data *)arg; + lump = &lmdp->lmd_lmm; + } + if (copy_to_user(lump, lmm, lmmsize)) + GOTO(out_lmm, rc = -EFAULT); + skip_lmm: + if (cmd == IOC_MDC_GETFILEINFO || cmd == LL_IOC_MDC_GETINFO) { + struct lov_user_mds_data *lmdp; + lstat_t st = { 0 }; + + st.st_dev = inode->i_sb->s_dev; + st.st_mode = body->mode; + st.st_nlink = body->nlink; + st.st_uid = body->uid; + st.st_gid = body->gid; + st.st_rdev = body->rdev; + st.st_size = body->size; + st.st_blksize = CFS_PAGE_SIZE; + st.st_blocks = body->blocks; + st.st_atime = body->atime; + st.st_mtime = body->mtime; + st.st_ctime = body->ctime; + st.st_ino = inode->i_ino; + + lmdp = (struct lov_user_mds_data *)arg; + if (copy_to_user(&lmdp->lmd_st, &st, sizeof(st))) + GOTO(out_lmm, rc = -EFAULT); + } - if (!IS_ERR(page)) { - de = ext2_next_entry((ext2_dirent *) page_address(page)); - *p = page; + EXIT; + out_lmm: + if (lmm && lmm->lmm_magic == LOV_MAGIC_JOIN) + OBD_FREE(lmm, lmmsize); + out_req: + ptlrpc_req_finished(request); + if (filename) + putname(filename); + return rc; } - return de; -} + case IOC_LOV_GETINFO: { + struct lov_user_mds_data *lumd; + struct lov_stripe_md *lsm; + struct lov_user_md *lum; + struct lov_mds_md *lmm; + int lmmsize; + lstat_t st; + int rc; + + lumd = (struct lov_user_mds_data *)arg; + lum = &lumd->lmd_lmm; + + rc = ll_get_max_mdsize(sbi, &lmmsize); + if (rc) + RETURN(rc); + + OBD_ALLOC(lmm, lmmsize); + if (copy_from_user(lmm, lum, lmmsize)) + GOTO(free_lmm, rc = -EFAULT); + + switch (lmm->lmm_magic) { + case LOV_USER_MAGIC_V1: + if (LOV_USER_MAGIC_V1 == cpu_to_le32(LOV_USER_MAGIC_V1)) + break; + /* swab objects first so that stripes num will be sane */ + lustre_swab_lov_user_md_objects( + ((struct lov_user_md_v1 *)lmm)->lmm_objects, + ((struct lov_user_md_v1 *)lmm)->lmm_stripe_count); + lustre_swab_lov_user_md_v1((struct lov_user_md_v1 *)lmm); + break; + case LOV_USER_MAGIC_V3: + if (LOV_USER_MAGIC_V3 == cpu_to_le32(LOV_USER_MAGIC_V3)) + break; + /* swab objects first so that stripes num will be sane */ + lustre_swab_lov_user_md_objects( + ((struct lov_user_md_v3 *)lmm)->lmm_objects, + ((struct lov_user_md_v3 *)lmm)->lmm_stripe_count); + lustre_swab_lov_user_md_v3((struct lov_user_md_v3 *)lmm); + break; + default: + GOTO(free_lmm, rc = -EINVAL); + } -obd_id ll_inode_by_name(struct inode * dir, struct dentry *dentry, int *type) -{ - obd_id res = 0; - struct ext2_dir_entry_2 * de; - struct page *page; + rc = obd_unpackmd(sbi->ll_dt_exp, &lsm, lmm, lmmsize); + if (rc < 0) + GOTO(free_lmm, rc = -ENOMEM); + + rc = obd_checkmd(sbi->ll_dt_exp, sbi->ll_md_exp, lsm); + if (rc) + GOTO(free_lsm, rc); + + /* Perform glimpse_size operation. */ + memset(&st, 0, sizeof(st)); - de = ext2_find_entry (dir, dentry, &page); - if (de) { - res = le32_to_cpu(de->inode); - *type = ll_dt2fmt[de->file_type]; - kunmap(page); - page_cache_release(page); + rc = ll_glimpse_ioctl(sbi, lsm, &st); + if (rc) + GOTO(free_lsm, rc); + + if (copy_to_user(&lumd->lmd_st, &st, sizeof(st))) + GOTO(free_lsm, rc = -EFAULT); + + EXIT; + free_lsm: + obd_free_memmd(sbi->ll_dt_exp, &lsm); + free_lmm: + OBD_FREE(lmm, lmmsize); + return rc; } - return res; -} + case OBD_IOC_LLOG_CATINFO: { + struct ptlrpc_request *req = NULL; + char *buf = NULL; + char *str; + int len = 0; + int rc; + + rc = obd_ioctl_getdata(&buf, &len, (void *)arg); + if (rc) + RETURN(rc); + data = (void *)buf; + + if (!data->ioc_inlbuf1) { + obd_ioctl_freedata(buf, len); + RETURN(-EINVAL); + } -/* Releases the page */ -void ext2_set_link(struct inode *dir, struct ext2_dir_entry_2 *de, - struct page *page, struct inode *inode) -{ - unsigned from = (char *) de - (char *) page_address(page); - unsigned to = from + le16_to_cpu(de->rec_len); - int err; + req = ptlrpc_request_alloc(sbi2mdc(sbi)->cl_import, + &RQF_LLOG_CATINFO); + if (req == NULL) + GOTO(out_catinfo, rc = -ENOMEM); - lock_page(page); - err = page->mapping->a_ops->prepare_write(NULL, page, from, to); - if (err) - LBUG(); - de->inode = cpu_to_le32(inode->i_ino); - ext2_set_de_type (de, inode); - dir->i_mtime = dir->i_ctime = CURRENT_TIME; - err = ext2_commit_chunk(page, from, to); - UnlockPage(page); - ext2_put_page(page); -} + req_capsule_set_size(&req->rq_pill, &RMF_NAME, RCL_CLIENT, + data->ioc_inllen1); + req_capsule_set_size(&req->rq_pill, &RMF_STRING, RCL_CLIENT, + data->ioc_inllen2); -/* - * Parent is locked. - */ -int ll_add_link (struct dentry *dentry, struct inode *inode) -{ - struct inode *dir = dentry->d_parent->d_inode; - const char *name = dentry->d_name.name; - int namelen = dentry->d_name.len; - unsigned reclen = EXT2_DIR_REC_LEN(namelen); - unsigned short rec_len, name_len; - struct page *page = NULL; - ext2_dirent * de; - unsigned long npages = dir_pages(dir); - unsigned long n; - char *kaddr; - unsigned from, to; - int err; - - /* We take care of directory expansion in the same loop */ - for (n = 0; n <= npages; n++) { - page = ext2_get_page(dir, n); - err = PTR_ERR(page); - if (IS_ERR(page)) - goto out; - kaddr = page_address(page); - de = (ext2_dirent *)kaddr; - kaddr += PAGE_CACHE_SIZE - reclen; - while ((char *)de <= kaddr) { - err = -EEXIST; - if (ext2_match (namelen, name, de)) - goto out_page; - name_len = EXT2_DIR_REC_LEN(de->name_len); - rec_len = le16_to_cpu(de->rec_len); - if ( n==npages && rec_len == 0) { - CERROR("Fatal dir behaviour\n"); - goto out_page; - } - if (!de->inode && rec_len >= reclen) - goto got_it; - if (rec_len >= name_len + reclen) - goto got_it; - de = (ext2_dirent *) ((char *) de + rec_len); + rc = ptlrpc_request_pack(req, LUSTRE_LOG_VERSION, LLOG_CATINFO); + if (rc) { + ptlrpc_request_free(req); + GOTO(out_catinfo, rc); } - ext2_put_page(page); + + str = req_capsule_client_get(&req->rq_pill, &RMF_NAME); + memcpy(str, data->ioc_inlbuf1, data->ioc_inllen1); + if (data->ioc_inllen2) { + str = req_capsule_client_get(&req->rq_pill, + &RMF_STRING); + memcpy(str, data->ioc_inlbuf2, data->ioc_inllen2); + } + + req_capsule_set_size(&req->rq_pill, &RMF_STRING, RCL_SERVER, + data->ioc_plen1); + ptlrpc_request_set_replen(req); + + rc = ptlrpc_queue_wait(req); + if (!rc) { + str = req_capsule_server_get(&req->rq_pill, + &RMF_STRING); + if (copy_to_user(data->ioc_pbuf1, str, data->ioc_plen1)) + rc = -EFAULT; + } + ptlrpc_req_finished(req); + out_catinfo: + obd_ioctl_freedata(buf, len); + RETURN(rc); } - LBUG(); - return -EINVAL; + case OBD_IOC_QUOTACHECK: { + struct obd_quotactl *oqctl; + int rc, error = 0; + + if (!cfs_capable(CFS_CAP_SYS_ADMIN) || + sbi->ll_flags & LL_SBI_RMT_CLIENT) + RETURN(-EPERM); + + OBD_ALLOC_PTR(oqctl); + if (!oqctl) + RETURN(-ENOMEM); + oqctl->qc_type = arg; + rc = obd_quotacheck(sbi->ll_md_exp, oqctl); + if (rc < 0) { + CDEBUG(D_INFO, "md_quotacheck failed: rc %d\n", rc); + error = rc; + } -got_it: - from = (char*)de - (char*)page_address(page); - to = from + rec_len; - lock_page(page); - err = page->mapping->a_ops->prepare_write(NULL, page, from, to); - if (err) - goto out_unlock; - if (de->inode) { - ext2_dirent *de1 = (ext2_dirent *) ((char *) de + name_len); - de1->rec_len = cpu_to_le16(rec_len - name_len); - de->rec_len = cpu_to_le16(name_len); - de = de1; + rc = obd_quotacheck(sbi->ll_dt_exp, oqctl); + if (rc < 0) + CDEBUG(D_INFO, "obd_quotacheck failed: rc %d\n", rc); + + OBD_FREE_PTR(oqctl); + return error ?: rc; + } + case OBD_IOC_POLL_QUOTACHECK: { + struct if_quotacheck *check; + int rc; + + if (!cfs_capable(CFS_CAP_SYS_ADMIN) || + sbi->ll_flags & LL_SBI_RMT_CLIENT) + RETURN(-EPERM); + + OBD_ALLOC_PTR(check); + if (!check) + RETURN(-ENOMEM); + + rc = obd_iocontrol(cmd, sbi->ll_md_exp, 0, (void *)check, + NULL); + if (rc) { + CDEBUG(D_QUOTA, "mdc ioctl %d failed: %d\n", cmd, rc); + if (copy_to_user((void *)arg, check, sizeof(*check))) + rc = -EFAULT; + GOTO(out_poll, rc); + } + + rc = obd_iocontrol(cmd, sbi->ll_dt_exp, 0, (void *)check, + NULL); + if (rc) { + CDEBUG(D_QUOTA, "osc ioctl %d failed: %d\n", cmd, rc); + if (copy_to_user((void *)arg, check, sizeof(*check))) + rc = -EFAULT; + GOTO(out_poll, rc); + } + out_poll: + OBD_FREE_PTR(check); + RETURN(rc); } - de->name_len = namelen; - memcpy (de->name, name, namelen); - de->inode = cpu_to_le32(inode->i_ino); - ext2_set_de_type (de, inode); - CDEBUG(D_INODE, "type set to %o\n", de->file_type); - dir->i_mtime = dir->i_ctime = CURRENT_TIME; - err = ext2_commit_chunk(page, from, to); + case OBD_IOC_QUOTACTL: { + struct if_quotactl *qctl; + int cmd, type, id, valid, rc = 0; + + OBD_ALLOC_PTR(qctl); + if (!qctl) + RETURN(-ENOMEM); + + if (copy_from_user(qctl, (void *)arg, sizeof(*qctl))) + GOTO(out_quotactl, rc = -EFAULT); + + cmd = qctl->qc_cmd; + type = qctl->qc_type; + id = qctl->qc_id; + valid = qctl->qc_valid; + + switch (cmd) { + case LUSTRE_Q_INVALIDATE: + case LUSTRE_Q_FINVALIDATE: + case Q_QUOTAON: + case Q_QUOTAOFF: + case Q_SETQUOTA: + case Q_SETINFO: + if (!cfs_capable(CFS_CAP_SYS_ADMIN) || + sbi->ll_flags & LL_SBI_RMT_CLIENT) + GOTO(out_quotactl, rc = -EPERM); + break; + case Q_GETQUOTA: + if (((type == USRQUOTA && current->euid != id) || + (type == GRPQUOTA && !in_egroup_p(id))) && + (!cfs_capable(CFS_CAP_SYS_ADMIN) || + sbi->ll_flags & LL_SBI_RMT_CLIENT)) + GOTO(out_quotactl, rc = -EPERM); + break; + case Q_GETINFO: + break; + default: + CERROR("unsupported quotactl op: %#x\n", cmd); + GOTO(out_quotactl, rc = -ENOTTY); + } - // change_inode happens with the commit_chunk - /* XXX OFFSET_CACHE */ + if (valid != QC_GENERAL) { + if (sbi->ll_flags & LL_SBI_RMT_CLIENT) + GOTO(out_quotactl, rc = -EOPNOTSUPP); + + if (cmd == Q_GETINFO) + qctl->qc_cmd = Q_GETOINFO; + else if (cmd == Q_GETQUOTA) + qctl->qc_cmd = Q_GETOQUOTA; + else + GOTO(out_quotactl, rc = -EINVAL); + + switch (valid) { + case QC_MDTIDX: + rc = obd_iocontrol(OBD_IOC_QUOTACTL, + sbi->ll_md_exp, + sizeof(*qctl), qctl, NULL); + break; + case QC_OSTIDX: + rc = obd_iocontrol(OBD_IOC_QUOTACTL, + sbi->ll_dt_exp, + sizeof(*qctl), qctl, NULL); + break; + case QC_UUID: + rc = obd_iocontrol(OBD_IOC_QUOTACTL, + sbi->ll_md_exp, + sizeof(*qctl), qctl, NULL); + if (rc == -EAGAIN) + rc = obd_iocontrol(OBD_IOC_QUOTACTL, + sbi->ll_dt_exp, + sizeof(*qctl), qctl, + NULL); + break; + default: + rc = -EINVAL; + break; + } -out_unlock: - UnlockPage(page); -out_page: - ext2_put_page(page); -out: - return err; -} + if (rc) + GOTO(out_quotactl, rc); + else + qctl->qc_cmd = cmd; + } else { + struct obd_quotactl *oqctl; + + OBD_ALLOC_PTR(oqctl); + if (!oqctl) + GOTO(out_quotactl, rc = -ENOMEM); + + QCTL_COPY(oqctl, qctl); + rc = obd_quotactl(sbi->ll_md_exp, oqctl); + if (rc) { + if (rc != -EALREADY && cmd == Q_QUOTAON) { + oqctl->qc_cmd = Q_QUOTAOFF; + obd_quotactl(sbi->ll_md_exp, oqctl); + } + OBD_FREE_PTR(oqctl); + GOTO(out_quotactl, rc); + } else { + QCTL_COPY(qctl, oqctl); + OBD_FREE_PTR(oqctl); + } + } -/* - * ext2_delete_entry deletes a directory entry by merging it with the - * previous entry. Page is up-to-date. Releases the page. - */ -int ext2_delete_entry (struct ext2_dir_entry_2 * dir, struct page * page ) -{ - struct address_space *mapping = page->mapping; - struct inode *inode = mapping->host; - char *kaddr = page_address(page); - unsigned from = ((char*)dir - kaddr) & ~(ext2_chunk_size(inode)-1); - unsigned to = ((char*)dir - kaddr) + le16_to_cpu(dir->rec_len); - ext2_dirent * pde = NULL; - ext2_dirent * de = (ext2_dirent *) (kaddr + from); - int err; - - while ((char*)de < (char*)dir) { - pde = de; - de = ext2_next_entry(de); + if (copy_to_user((void *)arg, qctl, sizeof(*qctl))) + rc = -EFAULT; + + out_quotactl: + OBD_FREE_PTR(qctl); + RETURN(rc); } - if (pde) - from = (char*)pde - (char*)page_address(page); - lock_page(page); - err = mapping->a_ops->prepare_write(NULL, page, from, to); - if (err) - LBUG(); - if (pde) - pde->rec_len = cpu_to_le16(to-from); - dir->inode = 0; - inode->i_ctime = inode->i_mtime = CURRENT_TIME; - err = ext2_commit_chunk(page, from, to); - UnlockPage(page); - ext2_put_page(page); - return err; -} + case OBD_IOC_GETNAME: { + struct obd_device *obd = class_exp2obd(sbi->ll_dt_exp); + if (!obd) + RETURN(-EFAULT); + if (copy_to_user((void *)arg, obd->obd_name, + strlen(obd->obd_name) + 1)) + RETURN (-EFAULT); + RETURN(0); + } + case LL_IOC_FLUSHCTX: + RETURN(ll_flush_ctx(inode)); +#ifdef CONFIG_FS_POSIX_ACL + case LL_IOC_RMTACL: { + if (sbi->ll_flags & LL_SBI_RMT_CLIENT && + inode == inode->i_sb->s_root->d_inode) { + struct ll_file_data *fd = LUSTRE_FPRIVATE(file); + int rc; + + LASSERT(fd != NULL); + rc = rct_add(&sbi->ll_rct, cfs_curproc_pid(), arg); + if (!rc) + fd->fd_flags |= LL_FILE_RMTACL; + RETURN(rc); + } else + RETURN(0); + } +#endif + case LL_IOC_GETOBDCOUNT: { + int count; + + if (copy_from_user(&count, (int *)arg, sizeof(int))) + RETURN(-EFAULT); + + if (!count) { + /* get ost count */ + struct lov_obd *lov = &sbi->ll_dt_exp->exp_obd->u.lov; + count = lov->desc.ld_tgt_count; + } else { + /* get mdt count */ + struct lmv_obd *lmv = &sbi->ll_md_exp->exp_obd->u.lmv; + count = lmv->desc.ld_tgt_count; + } -/* - * Set the first fragment of directory. - */ -int ext2_make_empty(struct inode *inode, struct inode *parent) -{ - struct address_space *mapping = inode->i_mapping; - struct page *page = grab_cache_page(mapping, 0); - unsigned chunk_size = ext2_chunk_size(inode); - struct ext2_dir_entry_2 * de; - char *base; - int err; - ENTRY; + if (copy_to_user((int *)arg, &count, sizeof(int))) + RETURN(-EFAULT); - if (!page) - return -ENOMEM; - base = kmap(page); - if (!base) - return -ENOMEM; + RETURN(0); + } + case LL_IOC_PATH2FID: + if (copy_to_user((void *)arg, &ll_i2info(inode)->lli_fid, + sizeof(struct lu_fid))) + RETURN(-EFAULT); + RETURN(0); + case OBD_IOC_CHANGELOG_CLEAR: { + struct ioc_changelog_clear *icc; + int rc; + + OBD_ALLOC_PTR(icc); + if (icc == NULL) + RETURN(-ENOMEM); + if (copy_from_user(icc, (void *)arg, sizeof(*icc))) + GOTO(icc_free, rc = -EFAULT); + + rc = obd_iocontrol(cmd, sbi->ll_md_exp, sizeof(*icc), icc,NULL); + +icc_free: + OBD_FREE_PTR(icc); + RETURN(rc); + } + case OBD_IOC_FID2PATH: + RETURN(ll_fid2path(ll_i2mdexp(inode), (void *)arg)); - err = mapping->a_ops->prepare_write(NULL, page, 0, chunk_size); - if (err) - goto fail; + default: + RETURN(obd_iocontrol(cmd, sbi->ll_dt_exp,0,NULL,(void *)arg)); + } +} - de = (struct ext2_dir_entry_2 *) base; - de->name_len = 1; - de->rec_len = cpu_to_le16(EXT2_DIR_REC_LEN(1)); - memcpy (de->name, ".\0\0", 4); - de->inode = cpu_to_le32(inode->i_ino); - ext2_set_de_type (de, inode); - - de = (struct ext2_dir_entry_2 *) (base + EXT2_DIR_REC_LEN(1)); - de->name_len = 2; - de->rec_len = cpu_to_le16(chunk_size - EXT2_DIR_REC_LEN(1)); - de->inode = cpu_to_le32(parent->i_ino); - memcpy (de->name, "..\0", 4); - ext2_set_de_type (de, inode); - - err = ext2_commit_chunk(page, 0, chunk_size); -fail: - kunmap(page); - UnlockPage(page); - page_cache_release(page); +int ll_dir_open(struct inode *inode, struct file *file) +{ ENTRY; - return err; + RETURN(ll_file_open(inode, file)); } -/* - * routine to check that the specified directory is empty (for rmdir) - */ -int ext2_empty_dir (struct inode * inode) +int ll_dir_release(struct inode *inode, struct file *file) { - struct page *page = NULL; - unsigned long i, npages = dir_pages(inode); - - for (i = 0; i < npages; i++) { - char *kaddr; - ext2_dirent * de; - page = ext2_get_page(inode, i); - - if (IS_ERR(page)) - continue; - - kaddr = page_address(page); - de = (ext2_dirent *)kaddr; - kaddr += PAGE_CACHE_SIZE-EXT2_DIR_REC_LEN(1); - - while ((char *)de <= kaddr) { - if (de->inode != 0) { - /* check for . and .. */ - if (de->name[0] != '.') - goto not_empty; - if (de->name_len > 2) - goto not_empty; - if (de->name_len < 2) { - if (de->inode != - cpu_to_le32(inode->i_ino)) - goto not_empty; - } else if (de->name[1] != '.') - goto not_empty; - } - de = ext2_next_entry(de); - } - ext2_put_page(page); - } - return 1; - -not_empty: - ext2_put_page(page); - return 0; + ENTRY; + RETURN(ll_file_release(inode, file)); } struct file_operations ll_dir_operations = { - read: generic_read_dir, - readdir: ll_readdir + .open = ll_dir_open, + .release = ll_dir_release, + .read = generic_read_dir, + .readdir = ll_readdir, + .ioctl = ll_dir_ioctl };