X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fllite%2Fdir.c;h=febf6ea759e574c1fc77d905d4958d99f6bc1a97;hp=6f26358f3b6af72f02285907173acb73577fd466;hb=86dc491a90fd3fcbd329d8844b7398a40e59644e;hpb=1cbf25caab01ac561bb35053ca34fb6eb7336b3d diff --git a/lustre/llite/dir.c b/lustre/llite/dir.c index 6f26358..febf6ea 100644 --- a/lustre/llite/dir.c +++ b/lustre/llite/dir.c @@ -1,6 +1,4 @@ -/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- - * vim:expandtab:shiftwidth=8:tabstop=8: - * +/* * GPL HEADER START * * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. @@ -28,6 +26,8 @@ /* * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. + * + * Copyright (c) 2011, 2013, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -42,12 +42,13 @@ #include #include #include -#include #include #include // for wait_on_buffer +#include #define DEBUG_SUBSYSTEM S_LLITE +#include #include #include #include @@ -57,15 +58,6 @@ #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 - /* * (new) readdir implementation overview. * @@ -138,84 +130,136 @@ * * page format * - * - * - * + * Page in MDS_READPAGE RPC is packed in LU_PAGE_SIZE, and each page contains + * a header lu_dirpage which describes the start/end hash, and whether this + * page is empty (contains no dir entry) or hash collide with next page. + * After client receives reply, several pages will be integrated into dir page + * in CFS_PAGE_SIZE (if CFS_PAGE_SIZE greater than LU_PAGE_SIZE), and the + * lu_dirpage for this integrated page will be adjusted. See + * lmv_adjust_dirpages(). * */ /* returns the page unlocked, but with a reference */ -static int ll_dir_readpage(struct file *file, struct page *page) +static int ll_dir_filler(void *_hash, struct page *page0) { - struct inode *inode = page->mapping->host; + struct inode *inode = page0->mapping->host; + int hash64 = ll_i2sbi(inode)->ll_flags & LL_SBI_64BIT_HASH; + struct obd_export *exp = ll_i2sbi(inode)->ll_md_exp; struct ptlrpc_request *request; struct mdt_body *body; - struct obd_capa *oc; - __u64 hash; + struct md_op_data *op_data; + __u64 hash = *((__u64 *)_hash); + struct page **page_pool; + struct page *page; +#ifndef HAVE_ADD_TO_PAGE_CACHE_LRU + struct pagevec lru_pvec; +#endif + struct lu_dirpage *dp; + int max_pages = ll_i2sbi(inode)->ll_md_brw_size >> CFS_PAGE_SHIFT; + int nrdpgs = 0; /* number of pages read actually */ + int npages; + int i; int rc; ENTRY; - if (file) { - struct ll_file_data *fd = LUSTRE_FPRIVATE(file); + CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p) hash "LPU64"\n", + inode->i_ino, inode->i_generation, inode, hash); - hash = fd->fd_dir.lfd_next; - } else { - struct ll_inode_info *lli = ll_i2info(inode); + LASSERT(max_pages > 0 && max_pages <= MD_MAX_BRW_PAGES); - cfs_spin_lock(&lli->lli_sa_lock); - if (lli->lli_sai) - LASSERT(lli->lli_sai->sai_pid == cfs_curproc_pid()); - else - LASSERT(lli->lli_opendir_pid == cfs_curproc_pid()); - hash = lli->lli_sa_pos; - cfs_spin_unlock(&lli->lli_sa_lock); + OBD_ALLOC(page_pool, sizeof(page) * max_pages); + if (page_pool != NULL) { + page_pool[0] = page0; + } else { + page_pool = &page0; + max_pages = 1; + } + for (npages = 1; npages < max_pages; npages++) { + page = page_cache_alloc_cold(inode->i_mapping); + if (!page) + break; + page_pool[npages] = page; } - 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) { + + op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL, 0, 0, + LUSTRE_OPC_ANY, NULL); + op_data->op_npages = npages; + op_data->op_offset = hash; + rc = md_readpage(exp, op_data, page_pool, &request); + ll_finish_md_op_data(op_data); + if (rc == 0) { body = req_capsule_server_get(&request->rq_pill, &RMF_MDT_BODY); /* Checked by mdc_readpage() */ LASSERT(body != NULL); if (body->valid & OBD_MD_FLSIZE) cl_isize_write(inode, body->size); - SetPageUptodate(page); + + nrdpgs = (request->rq_bulk->bd_nob_transferred+CFS_PAGE_SIZE-1) + >> CFS_PAGE_SHIFT; + SetPageUptodate(page0); } + unlock_page(page0); ptlrpc_req_finished(request); - unlock_page(page); + CDEBUG(D_VFSTRACE, "read %d/%d pages\n", nrdpgs, npages); + + ll_pagevec_init(&lru_pvec, 0); + for (i = 1; i < npages; i++) { + unsigned long offset; + int ret; + + page = page_pool[i]; + + if (rc < 0 || i >= nrdpgs) { + page_cache_release(page); + continue; + } + + SetPageUptodate(page); + + dp = cfs_kmap(page); + hash = le64_to_cpu(dp->ldp_hash_start); + cfs_kunmap(page); + + offset = hash_x_index(hash, hash64); + + prefetchw(&page->flags); + ret = add_to_page_cache_lru(page, inode->i_mapping, offset, + GFP_KERNEL); + if (ret == 0) { + unlock_page(page); + if (ll_pagevec_add(&lru_pvec, page) == 0) + ll_pagevec_lru_add_file(&lru_pvec); + } else { + CDEBUG(D_VFSTRACE, "page %lu add to page cache failed:" + " %d\n", offset, ret); + } + page_cache_release(page); + } + ll_pagevec_lru_add_file(&lru_pvec); + + if (page_pool != &page0) + OBD_FREE(page_pool, sizeof(struct page *) * max_pages); EXIT; return rc; } -struct address_space_operations ll_dir_aops = { - .readpage = ll_dir_readpage, -}; - static void ll_check_page(struct inode *dir, struct page *page) { /* XXX: check page format later */ SetPageChecked(page); } -static void ll_release_page(struct page *page, __u64 hash, - __u64 start, __u64 end) +void ll_release_page(struct page *page, int remove) { kunmap(page); - lock_page(page); - if (likely(page->mapping != NULL)) { - ll_truncate_complete_page(page); + if (remove) { + lock_page(page); + if (likely(page->mapping != NULL)) + truncate_complete_page(page->mapping, 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); } @@ -226,13 +270,14 @@ static void ll_release_page(struct page *page, __u64 hash, static struct page *ll_dir_page_locate(struct inode *dir, __u64 *hash, __u64 *start, __u64 *end) { + int hash64 = ll_i2sbi(dir)->ll_flags & LL_SBI_64BIT_HASH; 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(*hash); + unsigned long offset = hash_x_index(*hash, hash64); struct page *page; int found; @@ -250,23 +295,35 @@ static struct page *ll_dir_page_locate(struct inode *dir, __u64 *hash, * hence, can avoid restart. * * In fact, page cannot be locked here at all, because - * ll_dir_readpage() does synchronous io. + * ll_dir_filler() does synchronous io. */ wait_on_page(page); if (PageUptodate(page)) { - dp = kmap(page); -#if BITS_PER_LONG == 32 - *start = le64_to_cpu(dp->ldp_hash_start) >> 32; - *end = le64_to_cpu(dp->ldp_hash_end) >> 32; - *hash = *hash >> 32; -#else - *start = le64_to_cpu(dp->ldp_hash_start); - *end = le64_to_cpu(dp->ldp_hash_end); -#endif + dp = cfs_kmap(page); + if (BITS_PER_LONG == 32 && hash64) { + *start = le64_to_cpu(dp->ldp_hash_start) >> 32; + *end = le64_to_cpu(dp->ldp_hash_end) >> 32; + *hash = *hash >> 32; + } else { + *start = le64_to_cpu(dp->ldp_hash_start); + *end = le64_to_cpu(dp->ldp_hash_end); + } LASSERTF(*start <= *hash, "start = "LPX64",end = " LPX64",hash = "LPX64"\n", *start, *end, *hash); - if (*hash > *end || (*end != *start && *hash == *end)) { - ll_release_page(page, *hash, *start, *end); + CDEBUG(D_VFSTRACE, "page %lu [%llu %llu], hash "LPU64"\n", + offset, *start, *end, *hash); + if (*hash > *end) { + ll_release_page(page, 0); + page = NULL; + } else if (*end != *start && *hash == *end) { + /* + * upon hash collision, remove this page, + * otherwise put page reference, and + * ll_get_dir_page() will issue RPC to fetch + * the page we want. + */ + ll_release_page(page, + le32_to_cpu(dp->ldp_flags) & LDF_COLLIDE); page = NULL; } } else { @@ -281,8 +338,8 @@ static struct page *ll_dir_page_locate(struct inode *dir, __u64 *hash, return page; } -struct page *ll_get_dir_page(struct file *filp, struct inode *dir, __u64 hash, - int exact, struct ll_dir_chain *chain) +struct page *ll_get_dir_page(struct inode *dir, __u64 hash, + struct ll_dir_chain *chain) { ldlm_policy_data_t policy = {.l_inodebits = {MDS_INODELOCK_UPDATE} }; struct address_space *mapping = dir->i_mapping; @@ -295,36 +352,48 @@ struct page *ll_get_dir_page(struct file *filp, struct inode *dir, __u64 hash, __u64 end = 0; __u64 lhash = hash; struct ll_inode_info *lli = ll_i2info(dir); + int hash64 = ll_i2sbi(dir)->ll_flags & LL_SBI_64BIT_HASH; 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: "DFID" at "LPU64": rc %d\n", - PFID(ll_inode2fid(dir)), hash, rc); - return ERR_PTR(rc); - } + if (!rc) { + struct ldlm_enqueue_info einfo = {.ei_type = LDLM_IBITS, + .ei_mode = mode, + .ei_cb_bl = + ll_md_blocking_ast, + .ei_cb_cp = + ldlm_completion_ast, + .ei_cb_gl = NULL, + .ei_cb_wg = NULL, + .ei_cbdata = NULL}; + 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: "DFID" at "LPU64": rc %d\n", + PFID(ll_inode2fid(dir)), hash, rc); + return ERR_PTR(rc); + } + + CDEBUG(D_INODE, "setting lr_lvb_inode to inode %p (%lu/%u)\n", + dir, dir->i_ino, dir->i_generation); + md_set_lock_data(ll_i2sbi(dir)->ll_md_exp, + &it.d.lustre.it_lock_handle, dir, NULL); } else { /* for cross-ref object, l_ast_data of the lock may not be set, * we reset it here */ @@ -333,15 +402,13 @@ struct page *ll_get_dir_page(struct file *filp, struct inode *dir, __u64 hash, } ldlm_lock_dump_handle(D_OTHER, &lockh); - cfs_down(&lli->lli_readdir_sem); + mutex_lock(&lli->lli_readdir_mutex); page = ll_dir_page_locate(dir, &lhash, &start, &end); if (IS_ERR(page)) { CERROR("dir page locate: "DFID" at "LPU64": rc %ld\n", PFID(ll_inode2fid(dir)), lhash, PTR_ERR(page)); GOTO(out_unlock, page); - } - - if (page != NULL) { + } else if (page != NULL) { /* * XXX nikita: not entirely correct handling of a corner case: * suppose hash chain of entries with hash value HASH crosses @@ -356,24 +423,11 @@ struct page *ll_get_dir_page(struct file *filp, struct inode *dir, __u64 hash, * it as an "overflow" page. 1. invalidate all pages at * once. 2. use HASH|1 as an index for P1. */ - if (exact && lhash != 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. - */ - CDEBUG(D_OTHER, "Stale readpage page %p: " - "start = "LPX64",end = "LPX64"hash ="LPX64"\n", - page, start, end, lhash); - ll_release_page(page, lhash, start, end); - } else { - GOTO(hash_collision, page); - } + GOTO(hash_collision, page); } - page = read_cache_page(mapping, hash_x_index(hash), - (filler_t*)mapping->a_ops->readpage, filp); + page = read_cache_page(mapping, hash_x_index(hash, hash64), + ll_dir_filler, &lhash); if (IS_ERR(page)) { CERROR("read cache page: "DFID" at "LPU64": rc %ld\n", PFID(ll_inode2fid(dir)), hash, PTR_ERR(page)); @@ -396,23 +450,23 @@ struct page *ll_get_dir_page(struct file *filp, struct inode *dir, __u64 hash, } hash_collision: dp = page_address(page); -#if BITS_PER_LONG == 32 - start = le64_to_cpu(dp->ldp_hash_start) >> 32; - end = le64_to_cpu(dp->ldp_hash_end) >> 32; - lhash = hash >> 32; -#else - start = le64_to_cpu(dp->ldp_hash_start); - end = le64_to_cpu(dp->ldp_hash_end); - lhash = hash; -#endif + if (BITS_PER_LONG == 32 && hash64) { + start = le64_to_cpu(dp->ldp_hash_start) >> 32; + end = le64_to_cpu(dp->ldp_hash_end) >> 32; + lhash = hash >> 32; + } else { + start = le64_to_cpu(dp->ldp_hash_start); + end = le64_to_cpu(dp->ldp_hash_end); + lhash = hash; + } if (end == start) { LASSERT(start == lhash); CWARN("Page-wide hash collision: "LPU64"\n", end); -#if BITS_PER_LONG == 32 - CWARN("Real page-wide hash collision at ["LPU64" "LPU64"] with " - "hash "LPU64"\n", le64_to_cpu(dp->ldp_hash_start), - le64_to_cpu(dp->ldp_hash_end), hash); -#endif + if (BITS_PER_LONG == 32 && hash64) + CWARN("Real page-wide hash collision at ["LPU64" "LPU64 + "] with hash "LPU64"\n", + le64_to_cpu(dp->ldp_hash_start), + le64_to_cpu(dp->ldp_hash_end), hash); /* * Fetch whole overflow chain... * @@ -421,49 +475,33 @@ hash_collision: goto fail; } out_unlock: - cfs_up(&lli->lli_readdir_sem); + mutex_unlock(&lli->lli_readdir_mutex); ldlm_lock_decref(&lockh, mode); return page; fail: - ll_put_page(page); + ll_release_page(page, 1); page = ERR_PTR(-EIO); goto out_unlock; } -int ll_readdir(struct file *filp, void *cookie, filldir_t filldir) +int ll_dir_read(struct inode *inode, __u64 *_pos, void *cookie, + filldir_t filldir) { - struct inode *inode = filp->f_dentry->d_inode; - struct ll_inode_info *info = ll_i2info(inode); - struct ll_sb_info *sbi = ll_i2sbi(inode); - struct ll_file_data *fd = LUSTRE_FPRIVATE(filp); - __u64 pos = fd->fd_dir.lfd_pos; + struct ll_inode_info *info = ll_i2info(inode); + struct ll_sb_info *sbi = ll_i2sbi(inode); + __u64 pos = *_pos; + int api32 = ll_need_32bit_api(sbi); + int hash64 = sbi->ll_flags & LL_SBI_64BIT_HASH; struct page *page; struct ll_dir_chain chain; - int rc, need_32bit; - int done; - int shift; - __u16 type; + int done = 0; + int rc = 0; ENTRY; - need_32bit = ll_need_32bit_api(sbi); - CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p) pos %lu/%llu 32bit_api %d\n", - inode->i_ino, inode->i_generation, inode, - (unsigned long)pos, i_size_read(inode), need_32bit); - - if (pos == DIR_END_OFF) - /* - * end-of-file. - */ - RETURN(0); - - rc = 0; - done = 0; - shift = 0; ll_dir_chain_init(&chain); - fd->fd_dir.lfd_next = pos; - page = ll_get_dir_page(filp, inode, pos, 0, &chain); + page = ll_get_dir_page(inode, pos, &chain); while (rc == 0 && !done) { struct lu_dirpage *dp; @@ -474,17 +512,17 @@ int ll_readdir(struct file *filp, void *cookie, filldir_t filldir) * If page is empty (end of directory is reached), * use this value. */ - __u64 hash = DIR_END_OFF; + __u64 hash = MDS_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; + __u16 type; int namelen; struct lu_fid fid; - __u64 ino; __u64 lhash; + __u64 ino; /* * XXX: implement correct swabbing here. @@ -505,43 +543,51 @@ int ll_readdir(struct file *filp, void *cookie, filldir_t filldir) */ continue; - name = ent->lde_name; - fid_le_to_cpu(&fid, &ent->lde_fid); - if (need_32bit) { + if (api32 && hash64) lhash = hash >> 32; - ino = cl_fid_build_ino32(&fid); - } else { + else lhash = hash; - ino = cl_fid_build_ino(&fid); - } + fid_le_to_cpu(&fid, &ent->lde_fid); + ino = cl_fid_build_ino(&fid, api32); type = ll_dirent_type_get(ent); - done = filldir(cookie, name, namelen, + /* For 'll_nfs_get_name_filldir()', it will try + * to access the 'ent' through its 'lde_name', + * so the parameter 'name' for 'filldir()' must + * be part of the 'ent'. */ + done = filldir(cookie, ent->lde_name, namelen, lhash, ino, type); } next = le64_to_cpu(dp->ldp_hash_end); - ll_put_page(page); if (!done) { pos = next; - if (pos == DIR_END_OFF) { + if (pos == MDS_DIR_END_OFF) { /* * End of directory reached. */ done = 1; + ll_release_page(page, 0); } else if (1 /* chain is exhausted*/) { /* * Normal case: continue to the next * page. */ - fd->fd_dir.lfd_next = pos; - page = ll_get_dir_page(filp, inode, pos, - 1, &chain); + ll_release_page(page, + le32_to_cpu(dp->ldp_flags) & + LDF_COLLIDE); + next = pos; + page = ll_get_dir_page(inode, pos, + &chain); } else { /* * go into overflow page. */ + LASSERT(le32_to_cpu(dp->ldp_flags) & + LDF_COLLIDE); + ll_release_page(page, 1); } } else { pos = hash; + ll_release_page(page, 0); } } else { rc = PTR_ERR(page); @@ -550,17 +596,62 @@ int ll_readdir(struct file *filp, void *cookie, filldir_t filldir) } } - fd->fd_dir.lfd_pos = pos; - if (need_32bit) - filp->f_pos = pos >> 32; - else - filp->f_pos = pos; - filp->f_version = inode->i_version; - touch_atime(filp->f_vfsmnt, filp->f_dentry); + *_pos = pos; + ll_dir_chain_fini(&chain); + RETURN(rc); +} + +static int ll_readdir(struct file *filp, void *cookie, filldir_t filldir) +{ + struct inode *inode = filp->f_dentry->d_inode; + struct ll_file_data *lfd = LUSTRE_FPRIVATE(filp); + struct ll_sb_info *sbi = ll_i2sbi(inode); + __u64 pos = lfd->lfd_pos; + int hash64 = sbi->ll_flags & LL_SBI_64BIT_HASH; + int api32 = ll_need_32bit_api(sbi); + int rc; +#ifdef HAVE_TOUCH_ATIME_1ARG + struct path path; +#endif + ENTRY; + + CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p) pos %lu/%llu " + " 32bit_api %d\n", inode->i_ino, inode->i_generation, + inode, (unsigned long)pos, i_size_read(inode), api32); + + if (pos == MDS_DIR_END_OFF) + /* + * end-of-file. + */ + GOTO(out, rc = 0); + + rc = ll_dir_read(inode, &pos, cookie, filldir); + lfd->lfd_pos = pos; + if (pos == MDS_DIR_END_OFF) { + if (api32) + filp->f_pos = LL_DIR_END_OFF_32BIT; + else + filp->f_pos = LL_DIR_END_OFF; + } else { + if (api32 && hash64) + filp->f_pos = pos >> 32; + else + filp->f_pos = pos; + } + filp->f_version = inode->i_version; +#ifdef HAVE_TOUCH_ATIME_1ARG + path.mnt = filp->f_vfsmnt; + path.dentry = filp->f_dentry; + touch_atime(&path); +#else + touch_atime(filp->f_vfsmnt, filp->f_dentry); +#endif - ll_dir_chain_fini(&chain); +out: + if (!rc) + ll_stats_ops_tally(sbi, LPROC_LL_READDIR, 1); - RETURN(rc); + RETURN(rc); } int ll_send_mgc_param(struct obd_export *mgc, char *string) @@ -573,7 +664,7 @@ int ll_send_mgc_param(struct obd_export *mgc, char *string) return -ENOMEM; strncpy(msp->mgs_param, string, MGS_PARAM_MAXLEN); - rc = obd_set_info_async(mgc, sizeof(KEY_SET_INFO), KEY_SET_INFO, + rc = obd_set_info_async(NULL, 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); @@ -582,21 +673,34 @@ int ll_send_mgc_param(struct obd_export *mgc, char *string) return rc; } -char *ll_get_fsname(struct inode *inode) +int ll_dir_setdirstripe(struct inode *dir, struct lmv_user_md *lump, + char *filename) { - 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; + struct ptlrpc_request *request = NULL; + struct md_op_data *op_data; + struct ll_sb_info *sbi = ll_i2sbi(dir); + int mode; + int err; + + ENTRY; + + mode = (0755 & (S_IRWXUGO|S_ISVTX) & ~current->fs->umask) | S_IFDIR; + op_data = ll_prep_md_op_data(NULL, dir, NULL, filename, + strlen(filename), mode, LUSTRE_OPC_MKDIR, + lump); + if (IS_ERR(op_data)) + GOTO(err_exit, err = PTR_ERR(op_data)); + + op_data->op_cli_flags |= CLI_SET_MEA; + err = md_create(sbi->ll_md_exp, op_data, lump, sizeof(*lump), mode, + cfs_curproc_fsuid(), cfs_curproc_fsgid(), + cfs_curproc_cap_pack(), 0, &request); + ll_finish_md_op_data(op_data); + if (err) + GOTO(err_exit, err); +err_exit: + ptlrpc_req_finished(request); + return err; } int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump, @@ -608,8 +712,8 @@ int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump, 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; + ENTRY; if (lump != NULL) { /* @@ -623,22 +727,22 @@ int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump, 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: { + } + 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); - } - } + } + } } else { lum_size = sizeof(struct lov_user_md_v1); } @@ -648,6 +752,9 @@ int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump, if (IS_ERR(op_data)) RETURN(PTR_ERR(op_data)); + if (lump != NULL && lump->lmm_magic == cpu_to_le32(LMV_USER_MAGIC)) + op_data->op_cli_flags |= CLI_SET_MEA; + /* 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); @@ -662,38 +769,44 @@ int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump, 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 ? le32_to_cpu(lump->lmm_stripe_size) : 0); - rc = ll_send_mgc_param(mgc->u.cli.cl_mgc_mgsexp, param); - if (rc) - goto end; + char *param = NULL; + char *buf; + + OBD_ALLOC(param, MGS_PARAM_MAXLEN); + if (param == NULL) + GOTO(end, rc = -ENOMEM); + + buf = param; + /* Get fsname and assume devname to be -MDT0000. */ + ll_get_fsname(inode->i_sb, buf, MTI_NAME_MAXLEN); + strcat(buf, "-MDT0000.lov"); + buf += strlen(buf); + + /* Set root stripesize */ + sprintf(buf, ".stripesize=%u", + lump ? le32_to_cpu(lump->lmm_stripe_size) : 0); + rc = ll_send_mgc_param(mgc->u.cli.cl_mgc_mgsexp, param); + if (rc) + GOTO(end, rc); + + /* Set root stripecount */ + sprintf(buf, ".stripecount=%hd", + lump ? le16_to_cpu(lump->lmm_stripe_count) : 0); + rc = ll_send_mgc_param(mgc->u.cli.cl_mgc_mgsexp, param); + if (rc) + GOTO(end, rc); + + /* Set root stripeoffset */ + sprintf(buf, ".stripeoffset=%hd", + lump ? le16_to_cpu(lump->lmm_stripe_offset) : + (typeof(lump->lmm_stripe_offset))(-1)); + rc = ll_send_mgc_param(mgc->u.cli.cl_mgc_mgsexp, param); - /* Set root stripecount */ - sprintf(param, "%s-MDT0000.lov.stripecount=%hd", fsname, - lump ? le16_to_cpu(lump->lmm_stripe_count) : 0); - 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 ? le16_to_cpu(lump->lmm_stripe_offset) : - (typeof(lump->lmm_stripe_offset))(-1)); - 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; + if (param != NULL) + OBD_FREE(param, MGS_PARAM_MAXLEN); + } + RETURN(rc); } int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmmp, @@ -713,8 +826,8 @@ int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmmp, op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL, 0, lmmsize, LUSTRE_OPC_ANY, NULL); - if (op_data == NULL) - RETURN(-ENOMEM); + if (IS_ERR(op_data)) + RETURN(PTR_ERR(op_data)); op_data->op_valid = OBD_MD_FLEASIZE | OBD_MD_FLDIREA; rc = md_getattr(sbi->ll_md_exp, op_data, &req); @@ -778,10 +891,10 @@ int ll_get_mdt_idx(struct inode *inode) op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL, 0, 0, LUSTRE_OPC_ANY, NULL); - if (op_data == NULL) - RETURN(-ENOMEM); + if (IS_ERR(op_data)) + RETURN(PTR_ERR(op_data)); - op_data->op_valid |= OBD_MD_MDTIDX; + op_data->op_flags |= MF_GET_MDT_IDX; rc = md_getattr(sbi->ll_md_exp, op_data, NULL); mdtidx = op_data->op_mds; ll_finish_md_op_data(op_data); @@ -792,6 +905,176 @@ int ll_get_mdt_idx(struct inode *inode) return mdtidx; } +/** + * Generic handler to do any pre-copy work. + * + * It send a first hsm_progress (with extent length == 0) to coordinator as a + * first information for it that real work has started. + * + * Moreover, for a ARCHIVE request, it will sample the file data version and + * store it in \a copy. + * + * \return 0 on success. + */ +static int ll_ioc_copy_start(struct super_block *sb, struct hsm_copy *copy) +{ + struct ll_sb_info *sbi = ll_s2sbi(sb); + struct hsm_progress_kernel hpk; + int rc; + ENTRY; + + /* Forge a hsm_progress based on data from copy. */ + hpk.hpk_fid = copy->hc_hai.hai_fid; + hpk.hpk_cookie = copy->hc_hai.hai_cookie; + hpk.hpk_extent.offset = copy->hc_hai.hai_extent.offset; + hpk.hpk_extent.length = 0; + hpk.hpk_flags = 0; + hpk.hpk_errval = 0; + hpk.hpk_data_version = 0; + + + /* For archive request, we need to read the current file version. */ + if (copy->hc_hai.hai_action == HSMA_ARCHIVE) { + struct inode *inode; + __u64 data_version = 0; + + /* Get inode for this fid */ + inode = search_inode_for_lustre(sb, ©->hc_hai.hai_fid); + if (IS_ERR(inode)) { + hpk.hpk_flags |= HP_FLAG_RETRY; + /* hpk_errval is >= 0 */ + hpk.hpk_errval = -PTR_ERR(inode); + GOTO(progress, rc = PTR_ERR(inode)); + } + + /* Read current file data version */ + rc = ll_data_version(inode, &data_version, 1); + iput(inode); + if (rc != 0) { + CDEBUG(D_HSM, "Could not read file data version of " + DFID" (rc = %d). Archive request (" + LPX64") could not be done.\n", + PFID(©->hc_hai.hai_fid), rc, + copy->hc_hai.hai_cookie); + hpk.hpk_flags |= HP_FLAG_RETRY; + /* hpk_errval must be >= 0 */ + hpk.hpk_errval = -rc; + GOTO(progress, rc); + } + + /* Store it the hsm_copy for later copytool use. + * Always modified even if no lsm. */ + copy->hc_data_version = data_version; + } + +progress: + rc = obd_iocontrol(LL_IOC_HSM_PROGRESS, sbi->ll_md_exp, sizeof(hpk), + &hpk, NULL); + + RETURN(rc); +} + +/** + * Generic handler to do any post-copy work. + * + * It will send the last hsm_progress update to coordinator to inform it + * that copy is finished and whether it was successful or not. + * + * Moreover, + * - for ARCHIVE request, it will sample the file data version and compare it + * with the version saved in ll_ioc_copy_start(). If they do not match, copy + * will be considered as failed. + * - for RESTORE request, it will sample the file data version and send it to + * coordinator which is useful if the file was imported as 'released'. + * + * \return 0 on success. + */ +static int ll_ioc_copy_end(struct super_block *sb, struct hsm_copy *copy) +{ + struct ll_sb_info *sbi = ll_s2sbi(sb); + struct hsm_progress_kernel hpk; + int rc; + ENTRY; + + /* If you modify the logic here, also check llapi_hsm_copy_end(). */ + /* Take care: copy->hc_hai.hai_action, len, gid and data are not + * initialized if copy_end was called with copy == NULL. + */ + + /* Forge a hsm_progress based on data from copy. */ + hpk.hpk_fid = copy->hc_hai.hai_fid; + hpk.hpk_cookie = copy->hc_hai.hai_cookie; + hpk.hpk_extent = copy->hc_hai.hai_extent; + hpk.hpk_flags = copy->hc_flags | HP_FLAG_COMPLETED; + hpk.hpk_errval = copy->hc_errval; + hpk.hpk_data_version = 0; + + /* For archive request, we need to check the file data was not changed. + * + * For restore request, we need to send the file data version, this is + * useful when the file was created using hsm_import. + */ + if (((copy->hc_hai.hai_action == HSMA_ARCHIVE) || + (copy->hc_hai.hai_action == HSMA_RESTORE)) && + (copy->hc_errval == 0)) { + struct inode *inode; + __u64 data_version = 0; + + /* Get lsm for this fid */ + inode = search_inode_for_lustre(sb, ©->hc_hai.hai_fid); + if (IS_ERR(inode)) { + hpk.hpk_flags |= HP_FLAG_RETRY; + /* hpk_errval must be >= 0 */ + hpk.hpk_errval = -PTR_ERR(inode); + GOTO(progress, rc = PTR_ERR(inode)); + } + + rc = ll_data_version(inode, &data_version, + copy->hc_hai.hai_action == HSMA_ARCHIVE); + iput(inode); + if (rc) { + CDEBUG(D_HSM, "Could not read file data version. " + "Request could not be confirmed.\n"); + if (hpk.hpk_errval == 0) + hpk.hpk_errval = -rc; + GOTO(progress, rc); + } + + /* Store it the hsm_copy for later copytool use. + * Always modified even if no lsm. */ + hpk.hpk_data_version = data_version; + + /* File could have been stripped during archiving, so we need + * to check anyway. */ + if ((copy->hc_hai.hai_action == HSMA_ARCHIVE) && + (copy->hc_data_version != data_version)) { + CDEBUG(D_HSM, "File data version mismatched. " + "File content was changed during archiving. " + DFID", start:"LPX64" current:"LPX64"\n", + PFID(©->hc_hai.hai_fid), + copy->hc_data_version, data_version); + /* File was changed, send error to cdt. Do not ask for + * retry because if a file is modified frequently, + * the cdt will loop on retried archive requests. + * The policy engine will ask for a new archive later + * when the file will not be modified for some tunable + * time */ + /* we do not notify caller */ + hpk.hpk_flags &= ~HP_FLAG_RETRY; + /* hpk_errval must be >= 0 */ + hpk.hpk_errval = EBUSY; + } + + } + +progress: + rc = obd_iocontrol(LL_IOC_HSM_PROGRESS, sbi->ll_md_exp, sizeof(hpk), + &hpk, NULL); + + RETURN(rc); +} + + static int copy_and_ioctl(int cmd, struct obd_export *exp, void *data, int len) { void *ptr; @@ -809,9 +1092,145 @@ static int copy_and_ioctl(int cmd, struct obd_export *exp, void *data, int len) return rc; } -static int ll_dir_ioctl(struct inode *inode, struct file *file, - unsigned int cmd, unsigned long arg) +static int quotactl_ioctl(struct ll_sb_info *sbi, struct if_quotactl *qctl) { + int cmd = qctl->qc_cmd; + int type = qctl->qc_type; + int id = qctl->qc_id; + int valid = qctl->qc_valid; + int rc = 0; + ENTRY; + + 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) + RETURN(-EPERM); + break; + case Q_GETQUOTA: + if (((type == USRQUOTA && cfs_curproc_euid() != id) || + (type == GRPQUOTA && !in_egroup_p(id))) && + (!cfs_capable(CFS_CAP_SYS_ADMIN) || + sbi->ll_flags & LL_SBI_RMT_CLIENT)) + RETURN(-EPERM); + break; + case Q_GETINFO: + break; + default: + CERROR("unsupported quotactl op: %#x\n", cmd); + RETURN(-ENOTTY); + } + + if (valid != QC_GENERAL) { + if (sbi->ll_flags & LL_SBI_RMT_CLIENT) + RETURN(-EOPNOTSUPP); + + if (cmd == Q_GETINFO) + qctl->qc_cmd = Q_GETOINFO; + else if (cmd == Q_GETQUOTA) + qctl->qc_cmd = Q_GETOQUOTA; + else + RETURN(-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; + } + + if (rc) + RETURN(rc); + + qctl->qc_cmd = cmd; + } else { + struct obd_quotactl *oqctl; + + OBD_ALLOC_PTR(oqctl); + if (oqctl == NULL) + RETURN(-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); + RETURN(rc); + } + /* If QIF_SPACE is not set, client should collect the + * space usage from OSSs by itself */ + if (cmd == Q_GETQUOTA && + !(oqctl->qc_dqblk.dqb_valid & QIF_SPACE) && + !oqctl->qc_dqblk.dqb_curspace) { + struct obd_quotactl *oqctl_tmp; + + OBD_ALLOC_PTR(oqctl_tmp); + if (oqctl_tmp == NULL) + GOTO(out, rc = -ENOMEM); + + oqctl_tmp->qc_cmd = Q_GETOQUOTA; + oqctl_tmp->qc_id = oqctl->qc_id; + oqctl_tmp->qc_type = oqctl->qc_type; + + /* collect space usage from OSTs */ + oqctl_tmp->qc_dqblk.dqb_curspace = 0; + rc = obd_quotactl(sbi->ll_dt_exp, oqctl_tmp); + if (!rc || rc == -EREMOTEIO) { + oqctl->qc_dqblk.dqb_curspace = + oqctl_tmp->qc_dqblk.dqb_curspace; + oqctl->qc_dqblk.dqb_valid |= QIF_SPACE; + } + + /* collect space & inode usage from MDTs */ + oqctl_tmp->qc_dqblk.dqb_curspace = 0; + oqctl_tmp->qc_dqblk.dqb_curinodes = 0; + rc = obd_quotactl(sbi->ll_md_exp, oqctl_tmp); + if (!rc || rc == -EREMOTEIO) { + oqctl->qc_dqblk.dqb_curspace += + oqctl_tmp->qc_dqblk.dqb_curspace; + oqctl->qc_dqblk.dqb_curinodes = + oqctl_tmp->qc_dqblk.dqb_curinodes; + oqctl->qc_dqblk.dqb_valid |= QIF_INODES; + } else { + oqctl->qc_dqblk.dqb_valid &= ~QIF_SPACE; + } + + OBD_FREE_PTR(oqctl_tmp); + } +out: + QCTL_COPY(qctl, oqctl); + OBD_FREE_PTR(oqctl); + } + + RETURN(rc); +} + +static long ll_dir_ioctl(struct file *file, unsigned int cmd, unsigned long arg) +{ + struct inode *inode = file->f_dentry->d_inode; struct ll_sb_info *sbi = ll_i2sbi(inode); struct obd_ioctl_data *data; int rc = 0; @@ -872,8 +1291,8 @@ static int ll_dir_ioctl(struct inode *inode, struct file *file, op_data = ll_prep_md_op_data(NULL, inode, NULL, filename, namelen, 0, LUSTRE_OPC_ANY, NULL); - if (op_data == NULL) - GOTO(out_free, rc = -ENOMEM); + if (IS_ERR(op_data)) + GOTO(out_free, rc = PTR_ERR(op_data)); op_data->op_valid = OBD_MD_FLID; rc = md_getattr_name(sbi->ll_md_exp, op_data, &request); @@ -888,6 +1307,51 @@ out_free: obd_ioctl_freedata(buf, len); return rc; } + case LL_IOC_LMV_SETSTRIPE: { + struct lmv_user_md *lum; + char *buf = NULL; + char *filename; + int namelen = 0; + int lumlen = 0; + int len; + int rc; + + rc = obd_ioctl_getdata(&buf, &len, (void *)arg); + if (rc) + RETURN(rc); + + data = (void *)buf; + if (data->ioc_inlbuf1 == NULL || data->ioc_inlbuf2 == NULL || + data->ioc_inllen1 == 0 || data->ioc_inllen2 == 0) + GOTO(lmv_out_free, rc = -EINVAL); + + filename = data->ioc_inlbuf1; + namelen = data->ioc_inllen1; + + if (namelen < 1) { + CDEBUG(D_INFO, "IOC_MDC_LOOKUP missing filename\n"); + GOTO(lmv_out_free, rc = -EINVAL); + } + lum = (struct lmv_user_md *)data->ioc_inlbuf2; + lumlen = data->ioc_inllen2; + + if (lum->lum_magic != LMV_USER_MAGIC || + lumlen != sizeof(*lum)) { + CERROR("%s: wrong lum magic %x or size %d: rc = %d\n", + filename, lum->lum_magic, lumlen, -EFAULT); + GOTO(lmv_out_free, rc = -EINVAL); + } + + /** + * ll_dir_setdirstripe will be used to set dir stripe + * mdc_create--->mdt_reint_create (with dirstripe) + */ + rc = ll_dir_setdirstripe(inode, lum, filename); +lmv_out_free: + 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; @@ -903,7 +1367,7 @@ out_free: if (cfs_copy_from_user(lumv1, lumv1p, sizeof(*lumv1))) RETURN(-EFAULT); - if (lumv1->lmm_magic == LOV_USER_MAGIC_V3) { + if ((lumv1->lmm_magic == LOV_USER_MAGIC_V3) ) { if (cfs_copy_from_user(&lumv3, lumv3p, sizeof(lumv3))) RETURN(-EFAULT); } @@ -916,6 +1380,72 @@ out_free: RETURN(rc); } + case LL_IOC_LMV_GETSTRIPE: { + struct lmv_user_md *lump = (struct lmv_user_md *)arg; + struct lmv_user_md lum; + struct lmv_user_md *tmp; + int lum_size; + int rc = 0; + int mdtindex; + + if (copy_from_user(&lum, lump, sizeof(struct lmv_user_md))) + RETURN(-EFAULT); + + if (lum.lum_magic != LMV_MAGIC_V1) + RETURN(-EINVAL); + + lum_size = lmv_user_md_size(1, LMV_MAGIC_V1); + OBD_ALLOC(tmp, lum_size); + if (tmp == NULL) + GOTO(free_lmv, rc = -ENOMEM); + + memcpy(tmp, &lum, sizeof(lum)); + tmp->lum_type = LMV_STRIPE_TYPE; + tmp->lum_stripe_count = 1; + mdtindex = ll_get_mdt_idx(inode); + if (mdtindex < 0) + GOTO(free_lmv, rc = -ENOMEM); + + tmp->lum_stripe_offset = mdtindex; + tmp->lum_objects[0].lum_mds = mdtindex; + memcpy(&tmp->lum_objects[0].lum_fid, ll_inode2fid(inode), + sizeof(struct lu_fid)); + if (copy_to_user((void *)arg, tmp, lum_size)) + GOTO(free_lmv, rc = -EFAULT); +free_lmv: + if (tmp) + OBD_FREE(tmp, lum_size); + RETURN(rc); + } + case LL_IOC_REMOVE_ENTRY: { + char *filename = NULL; + int namelen = 0; + int rc; + + /* Here is a little hack to avoid sending REINT_RMENTRY to + * unsupported server, which might crash the server(LU-2730), + * Because both LVB_TYPE and REINT_RMENTRY will be supported + * on 2.4, we use OBD_CONNECT_LVB_TYPE to detect whether the + * server will support REINT_RMENTRY XXX*/ + if (!(exp_connect_flags(sbi->ll_md_exp) & OBD_CONNECT_LVB_TYPE)) + return -ENOTSUPP; + + filename = getname((const char *)arg); + if (IS_ERR(filename)) + RETURN(PTR_ERR(filename)); + + namelen = strlen(filename); + if (namelen < 1) + GOTO(out_rmdir, rc = -EINVAL); + + rc = ll_rmdir_entry(inode, filename, namelen); +out_rmdir: + if (filename) + putname(filename); + RETURN(rc); + } + case LL_IOC_LOV_SWAP_LAYOUTS: + RETURN(-EPERM); case LL_IOC_OBD_STATFS: RETURN(ll_obd_statfs(inode, (void *)arg)); case LL_IOC_LOV_GETSTRIPE: @@ -1016,9 +1546,12 @@ out_free: if (rc) RETURN(rc); - OBD_ALLOC_LARGE(lmm, lmmsize); - if (cfs_copy_from_user(lmm, lum, lmmsize)) - GOTO(free_lmm, rc = -EFAULT); + OBD_ALLOC_LARGE(lmm, lmmsize); + if (lmm == NULL) + RETURN(-ENOMEM); + + if (cfs_copy_from_user(lmm, lum, lmmsize)) + GOTO(free_lmm, rc = -EFAULT); switch (lmm->lmm_magic) { case LOV_USER_MAGIC_V1: @@ -1065,61 +1598,7 @@ out_free: return rc; } case OBD_IOC_LLOG_CATINFO: { - struct ptlrpc_request *req = NULL; - char *buf = NULL; - char *str; - int len = 0; - - 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); - } - - req = ptlrpc_request_alloc(sbi2mdc(sbi)->cl_import, - &RQF_LLOG_CATINFO); - if (req == NULL) - GOTO(out_catinfo, rc = -ENOMEM); - - 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); - - rc = ptlrpc_request_pack(req, LUSTRE_LOG_VERSION, LLOG_CATINFO); - if (rc) { - ptlrpc_request_free(req); - GOTO(out_catinfo, rc); - } - - 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 (cfs_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); + RETURN(-EOPNOTSUPP); } case OBD_IOC_QUOTACHECK: { struct obd_quotactl *oqctl; @@ -1180,9 +1659,62 @@ out_free: OBD_FREE_PTR(check); RETURN(rc); } - case OBD_IOC_QUOTACTL: { +#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) + case LL_IOC_QUOTACTL_18: { + /* copy the old 1.x quota struct for internal use, then copy + * back into old format struct. For 1.8 compatibility. */ + struct if_quotactl_18 *qctl_18; + struct if_quotactl *qctl_20; + + OBD_ALLOC_PTR(qctl_18); + if (!qctl_18) + RETURN(-ENOMEM); + + OBD_ALLOC_PTR(qctl_20); + if (!qctl_20) + GOTO(out_quotactl_18, rc = -ENOMEM); + + if (cfs_copy_from_user(qctl_18, (void *)arg, sizeof(*qctl_18))) + GOTO(out_quotactl_20, rc = -ENOMEM); + + QCTL_COPY(qctl_20, qctl_18); + qctl_20->qc_idx = 0; + + /* XXX: dqb_valid was borrowed as a flag to mark that + * only mds quota is wanted */ + if (qctl_18->qc_cmd == Q_GETQUOTA && + qctl_18->qc_dqblk.dqb_valid) { + qctl_20->qc_valid = QC_MDTIDX; + qctl_20->qc_dqblk.dqb_valid = 0; + } else if (qctl_18->obd_uuid.uuid[0] != '\0') { + qctl_20->qc_valid = QC_UUID; + qctl_20->obd_uuid = qctl_18->obd_uuid; + } else { + qctl_20->qc_valid = QC_GENERAL; + } + + rc = quotactl_ioctl(sbi, qctl_20); + + if (rc == 0) { + QCTL_COPY(qctl_18, qctl_20); + qctl_18->obd_uuid = qctl_20->obd_uuid; + + if (cfs_copy_to_user((void *)arg, qctl_18, + sizeof(*qctl_18))) + rc = -EFAULT; + } + + out_quotactl_20: + OBD_FREE_PTR(qctl_20); + out_quotactl_18: + OBD_FREE_PTR(qctl_18); + RETURN(rc); + } +#else +#warning "remove old LL_IOC_QUOTACTL_18 compatibility code" +#endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */ + case LL_IOC_QUOTACTL: { struct if_quotactl *qctl; - int cmd, type, id, valid; OBD_ALLOC_PTR(qctl); if (!qctl) @@ -1191,115 +1723,18 @@ out_free: if (cfs_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 && cfs_curproc_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); - } - - 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; - } - - 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); - } - } + rc = quotactl_ioctl(sbi, qctl); - if (cfs_copy_to_user((void *)arg, qctl, sizeof(*qctl))) + if (rc == 0 && cfs_copy_to_user((void *)arg,qctl,sizeof(*qctl))) rc = -EFAULT; out_quotactl: OBD_FREE_PTR(qctl); RETURN(rc); } - case OBD_IOC_GETNAME: { - struct obd_device *obd = class_exp2obd(sbi->ll_dt_exp); - if (!obd) - RETURN(-EFAULT); - if (cfs_copy_to_user((void *)arg, obd->obd_name, - strlen(obd->obd_name) + 1)) - RETURN (-EFAULT); - RETURN(0); - } + case OBD_IOC_GETDTNAME: + case OBD_IOC_GETMDNAME: + RETURN(ll_get_obd_name(inode, cmd, arg)); case LL_IOC_FLUSHCTX: RETURN(ll_flush_ctx(inode)); #ifdef CONFIG_FS_POSIX_ACL @@ -1318,19 +1753,20 @@ out_free: } #endif case LL_IOC_GETOBDCOUNT: { - int count; + int count, vallen; + struct obd_export *exp; if (cfs_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; + /* get ost count when count is zero, get mdt count otherwise */ + exp = count ? sbi->ll_md_exp : sbi->ll_dt_exp; + vallen = sizeof(count); + rc = obd_get_info(NULL, exp, sizeof(KEY_TGT_COUNT), + KEY_TGT_COUNT, &vallen, &count, NULL); + if (rc) { + CERROR("get target count failed: %d\n", rc); + RETURN(rc); } if (cfs_copy_to_user((int *)arg, &count, sizeof(int))) @@ -1352,46 +1788,159 @@ out_free: sizeof(struct ioc_changelog)); RETURN(rc); case OBD_IOC_FID2PATH: - RETURN(ll_fid2path(ll_i2mdexp(inode), (void *)arg)); - case LL_IOC_HSM_CT_START: - rc = copy_and_ioctl(cmd, sbi->ll_md_exp, (void *)arg, - sizeof(struct lustre_kernelcomm)); - RETURN(rc); - - default: - RETURN(obd_iocontrol(cmd, sbi->ll_dt_exp,0,NULL,(void *)arg)); - } + RETURN(ll_fid2path(inode, (void *)arg)); + case LL_IOC_HSM_REQUEST: { + struct hsm_user_request *hur; + int totalsize; + + OBD_ALLOC_PTR(hur); + if (hur == NULL) + RETURN(-ENOMEM); + + /* We don't know the true size yet; copy the fixed-size part */ + if (cfs_copy_from_user(hur, (void *)arg, sizeof(*hur))) { + OBD_FREE_PTR(hur); + RETURN(-EFAULT); + } + + /* Compute the whole struct size */ + totalsize = hur_len(hur); + OBD_FREE_PTR(hur); + OBD_ALLOC_LARGE(hur, totalsize); + if (hur == NULL) + RETURN(-ENOMEM); + + /* Copy the whole struct */ + if (cfs_copy_from_user(hur, (void *)arg, totalsize)) { + OBD_FREE_LARGE(hur, totalsize); + RETURN(-EFAULT); + } + + rc = obd_iocontrol(cmd, ll_i2mdexp(inode), totalsize, + hur, NULL); + + OBD_FREE_LARGE(hur, totalsize); + + RETURN(rc); + } + case LL_IOC_HSM_PROGRESS: { + struct hsm_progress_kernel hpk; + struct hsm_progress hp; + + if (cfs_copy_from_user(&hp, (void *)arg, sizeof(hp))) + RETURN(-EFAULT); + + hpk.hpk_fid = hp.hp_fid; + hpk.hpk_cookie = hp.hp_cookie; + hpk.hpk_extent = hp.hp_extent; + hpk.hpk_flags = hp.hp_flags; + hpk.hpk_errval = hp.hp_errval; + hpk.hpk_data_version = 0; + + /* File may not exist in Lustre; all progress + * reported to Lustre root */ + rc = obd_iocontrol(cmd, sbi->ll_md_exp, sizeof(hpk), &hpk, + NULL); + RETURN(rc); + } + case LL_IOC_HSM_CT_START: + rc = copy_and_ioctl(cmd, sbi->ll_md_exp, (void *)arg, + sizeof(struct lustre_kernelcomm)); + RETURN(rc); + + case LL_IOC_HSM_COPY_START: { + struct hsm_copy *copy; + int rc; + + OBD_ALLOC_PTR(copy); + if (copy == NULL) + RETURN(-ENOMEM); + if (cfs_copy_from_user(copy, (char *)arg, sizeof(*copy))) { + OBD_FREE_PTR(copy); + RETURN(-EFAULT); + } + + rc = ll_ioc_copy_start(inode->i_sb, copy); + if (cfs_copy_to_user((char *)arg, copy, sizeof(*copy))) + rc = -EFAULT; + + OBD_FREE_PTR(copy); + RETURN(rc); + } + case LL_IOC_HSM_COPY_END: { + struct hsm_copy *copy; + int rc; + + OBD_ALLOC_PTR(copy); + if (copy == NULL) + RETURN(-ENOMEM); + if (cfs_copy_from_user(copy, (char *)arg, sizeof(*copy))) { + OBD_FREE_PTR(copy); + RETURN(-EFAULT); + } + + rc = ll_ioc_copy_end(inode->i_sb, copy); + if (cfs_copy_to_user((char *)arg, copy, sizeof(*copy))) + rc = -EFAULT; + + OBD_FREE_PTR(copy); + RETURN(rc); + } + default: + RETURN(obd_iocontrol(cmd, sbi->ll_dt_exp, 0, NULL, + (void *)arg)); + } } static loff_t ll_dir_seek(struct file *file, loff_t offset, int origin) { + struct inode *inode = file->f_mapping->host; struct ll_file_data *fd = LUSTRE_FPRIVATE(file); - loff_t pos = file->f_pos; - loff_t ret; + struct ll_sb_info *sbi = ll_i2sbi(inode); + int api32 = ll_need_32bit_api(sbi); + loff_t ret = -EINVAL; ENTRY; - if (origin == 1 && offset >= 0 && file->f_pos == DIR_END_OFF) { - CWARN("end of dir hash, DIR_END_OFF(-2) is returned\n"); - RETURN(DIR_END_OFF); + mutex_lock(&inode->i_mutex); + switch (origin) { + case SEEK_SET: + break; + case SEEK_CUR: + offset += file->f_pos; + break; + case SEEK_END: + if (offset > 0) + GOTO(out, ret); + if (api32) + offset += LL_DIR_END_OFF_32BIT; + else + offset += LL_DIR_END_OFF; + break; + default: + GOTO(out, ret); } - ret = default_llseek(file, offset, origin); - if (ret >= 0) { - struct ll_sb_info *sbi = ll_i2sbi(file->f_dentry->d_inode); - - if (ll_need_32bit_api(sbi)) { - if (file->f_pos >> 32) { - /* hash overflow, simple revert */ - file->f_pos = pos; - RETURN(-EOVERFLOW); - } else { - fd->fd_dir.lfd_pos = file->f_pos << 32; - } - } else { - fd->fd_dir.lfd_pos = file->f_pos; + if (offset >= 0 && + ((api32 && offset <= LL_DIR_END_OFF_32BIT) || + (!api32 && offset <= LL_DIR_END_OFF))) { + if (offset != file->f_pos) { + if ((api32 && offset == LL_DIR_END_OFF_32BIT) || + (!api32 && offset == LL_DIR_END_OFF)) + fd->lfd_pos = MDS_DIR_END_OFF; + else if (api32 && sbi->ll_flags & LL_SBI_64BIT_HASH) + fd->lfd_pos = offset << 32; + else + fd->lfd_pos = offset; + file->f_pos = offset; + file->f_version = 0; } + ret = offset; } - RETURN(ret); + GOTO(out, ret); + +out: + mutex_unlock(&inode->i_mutex); + return ret; } int ll_dir_open(struct inode *inode, struct file *file) @@ -1412,5 +1961,6 @@ struct file_operations ll_dir_operations = { .release = ll_dir_release, .read = generic_read_dir, .readdir = ll_readdir, - .ioctl = ll_dir_ioctl + .unlocked_ioctl = ll_dir_ioctl, + .fsync = ll_fsync, };