X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Flvfs%2Ffsfilt_ext3.c;h=abdb74e2e5ee1812927b7c4f295d3fe7d28a87d8;hp=78ea86744a8a30833da45c57e6c807ed5b440423;hb=cf9e92a0adb490109f67d05da5833cfb270d92e8;hpb=e903932500fc08b143467ce5a1c2702df35d8f0f diff --git a/lustre/lvfs/fsfilt_ext3.c b/lustre/lvfs/fsfilt_ext3.c index 78ea867..abdb74e 100644 --- a/lustre/lvfs/fsfilt_ext3.c +++ b/lustre/lvfs/fsfilt_ext3.c @@ -1,26 +1,41 @@ /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- * vim:expandtab:shiftwidth=8:tabstop=8: * - * lustre/lib/fsfilt_ext3.c - * Lustre filesystem abstraction routines + * GPL HEADER START * - * Copyright (C) 2002, 2003 Cluster File Systems, Inc. - * Author: Andreas Dilger + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * - * This file is part of Lustre, http://www.lustre.org. + * 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. * - * Lustre is free software; you can redistribute it and/or - * modify it under the terms of version 2 of the GNU General Public - * License as published by the Free Software Foundation. + * 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). * - * Lustre 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 for more details. + * 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 * - * You should have received a copy of the GNU General Public License - * along with Lustre; if not, write to the Free Software - * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * 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. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2002, 2010, Oracle and/or its affiliates. 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. + * + * lustre/lvfs/fsfilt_ext3.c + * + * Author: Andreas Dilger */ #define DEBUG_SUBSYSTEM S_FILTER @@ -28,18 +43,37 @@ #include #include #include -#include #include #include #include +#ifdef HAVE_EXT4_LDISKFS +#include +#include +#else +#include #include #include +#endif #include #include #include -#include -#include +#ifdef HAVE_QUOTAIO_V1_H +# include +# include +#else +# include +# include +# include +# define V2_DQTREEOFF QT_TREEOFF +#endif + +#if defined(HAVE_EXT3_XATTR_H) #include +#else +/* ext3 xattr.h not available in rh style kernel-devel rpm */ +extern int ext3_xattr_get(struct inode *, int, const char *, void *, size_t); +extern int ext3_xattr_set_handle(handle_t *, struct inode *, int, const char *, const void *, size_t, int); +#endif #include #include @@ -48,10 +82,14 @@ #include #include -#ifdef EXT3_MULTIBLOCK_ALLOCATOR +#ifdef HAVE_EXT4_LDISKFS +#include +#else #include #endif +#include "lustre_quota_fmt.h" + #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,15) #define FSFILT_DATA_TRANS_BLOCKS(sb) EXT3_DATA_TRANS_BLOCKS #define FSFILT_DELETE_TRANS_BLOCKS(sb) EXT3_DELETE_TRANS_BLOCKS @@ -67,8 +105,14 @@ #define FSFILT_SINGLEDATA_TRANS_BLOCKS(sb) EXT3_SINGLEDATA_TRANS_BLOCKS #endif -#define fsfilt_ext3_journal_start(inode, nblocks) ext3_journal_start(inode, nblocks) -#define fsfilt_ext3_journal_stop(handle) ext3_journal_stop(handle) +#ifdef EXT_INSERT_EXTENT_WITH_5ARGS +#define fsfilt_ext3_ext_insert_extent(handle, inode, path, newext, flag) \ + ext3_ext_insert_extent(handle, inode, path, newext, flag) +#else +#define fsfilt_ext3_ext_insert_extent(handle, inode, path, newext, flag) \ + ext3_ext_insert_extent(handle, inode, path, newext) +#endif + static cfs_mem_cache_t *fcb_cache; @@ -84,6 +128,31 @@ struct fsfilt_cb_data { #define EXT3_XATTR_INDEX_TRUSTED 4 #endif +#ifdef HAVE_EXT4_LDISKFS +#define fsfilt_log_start_commit(journal, tid) jbd2_log_start_commit(journal, tid) +#define fsfilt_log_wait_commit(journal, tid) jbd2_log_wait_commit(journal, tid) +#define fsfilt_journal_callback_set(handle, func, jcb) jbd2_journal_callback_set(handle, func, jcb) +#else +#define fsfilt_log_start_commit(journal, tid) log_start_commit(journal, tid) +#define fsfilt_log_wait_commit(journal, tid) log_wait_commit(journal, tid) +#define fsfilt_journal_callback_set(handle, func, jcb) journal_callback_set(handle, func, jcb) +#define ext_pblock(ex) le32_to_cpu((ex)->ee_start) +#define ext3_ext_store_pblock(ex, pblock) ((ex)->ee_start = cpu_to_le32(pblock)) +#define ext3_inode_bitmap(sb,desc) le32_to_cpu((desc)->bg_inode_bitmap) +#endif + +#ifndef ext3_find_next_bit +#define ext3_find_next_bit ext2_find_next_bit +#endif + +#ifndef ext2_find_next_bit +#ifdef __LITTLE_ENDIAN +#define ext2_find_next_bit(addr, size, off) find_next_bit((unsigned long *)(addr), (size), (off)) +#else +error "Need implementation of find_next_bit on big-endian systems" +#endif /* __LITTLE_ENDIAN */ +#endif /* !ext2_find_next_le_bit */ + static char *fsfilt_ext3_get_label(struct super_block *sb) { return EXT3_SB(sb)->s_es->s_volume_name; @@ -97,7 +166,7 @@ static int fsfilt_ext3_set_label(struct super_block *sb, char *label) int err; journal = EXT3_SB(sb)->s_journal; - handle = journal_start(journal, 1); + handle = ext3_journal_start_sb(sb, 1); if (IS_ERR(handle)) { CERROR("can't start transaction\n"); return(PTR_ERR(handle)); @@ -113,7 +182,7 @@ static int fsfilt_ext3_set_label(struct super_block *sb, char *label) err = ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh); out: - journal_stop(handle); + ext3_journal_stop(handle); return(err); } @@ -124,12 +193,33 @@ static char *fsfilt_ext3_uuid(struct super_block *sb) } #ifdef HAVE_DISK_INODE_VERSION + +static __u64 get_i_version(struct inode *inode) +{ +#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27)) && defined(HAVE_EXT4_LDISKFS) + return inode->i_version; +#else + return EXT3_I(inode)->i_fs_version; +#endif +} + +static void set_i_version(struct inode *inode, __u64 new_version) +{ +#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27)) && defined(HAVE_EXT4_LDISKFS) + inode->i_version = new_version; +#else + (EXT3_I(inode))->i_fs_version = new_version; +#endif +} + /* * Get the 64-bit version for an inode. */ static __u64 fsfilt_ext3_get_version(struct inode *inode) { - return EXT3_I(inode)->i_fs_version; + CDEBUG(D_INFO, "Get version "LPX64" for inode %lu\n", + get_i_version(inode), inode->i_ino); + return get_i_version(inode); } /* @@ -137,9 +227,14 @@ static __u64 fsfilt_ext3_get_version(struct inode *inode) */ static __u64 fsfilt_ext3_set_version(struct inode *inode, __u64 new_version) { - __u64 old_version = EXT3_I(inode)->i_fs_version; - - (EXT3_I(inode))->i_fs_version = new_version; + __u64 old_version = get_i_version(inode); + + CDEBUG(D_INFO, "Set version "LPX64" (old "LPX64") for inode %lu\n", + new_version, old_version, inode->i_ino); + set_i_version(inode, new_version); + /* version is set after all inode operations are finished, so we should + * mark it dirty here */ + inode->i_sb->s_op->dirty_inode(inode); return old_version; } @@ -182,7 +277,7 @@ static void *fsfilt_ext3_start(struct inode *inode, int op, void *desc_private, nblocks += 3; /* no break */ case FSFILT_OP_CREATE: { -#if defined(EXT3_EXTENTS_FL) && defined(EXT3_INDEX_FL) +#if defined(EXT3_EXTENTS_FL) && defined(EXT3_INDEX_FL) && !defined(HAVE_EXT4_LDISKFS) static int warned; if (!warned) { if (!test_opt(inode->i_sb, EXTENTS)) { @@ -190,8 +285,9 @@ static void *fsfilt_ext3_start(struct inode *inode, int op, void *desc_private, } else if (((EXT3_I(inode)->i_flags & cpu_to_le32(EXT3_EXTENTS_FL | EXT3_INDEX_FL)) == cpu_to_le32(EXT3_EXTENTS_FL | EXT3_INDEX_FL))) { - CWARN("extent-mapped directory found - contact " - "CFS: support@clusterfs.com\n"); + CWARN("extent-mapped directory found with " + "ext3-based ldiskfs - contact " + "http://bugzilla.lustre.org/\n"); warned = 1; } } @@ -226,19 +322,6 @@ static void *fsfilt_ext3_start(struct inode *inode, int op, void *desc_private, nblocks = (LLOG_CHUNK_SIZE >> inode->i_blkbits) + FSFILT_DELETE_TRANS_BLOCKS(inode->i_sb) * logs; break; - case FSFILT_OP_JOIN: - /* delete 2 file(file + array id) + create 1 file (array id) - * create/update logs for each stripe */ - nblocks += 2 * FSFILT_DELETE_TRANS_BLOCKS(inode->i_sb); - - /*create array log for head file*/ - nblocks += 3; - nblocks += (EXT3_INDEX_EXTRA_TRANS_BLOCKS + - FSFILT_SINGLEDATA_TRANS_BLOCKS(inode->i_sb)); - /*update head file array */ - nblocks += EXT3_INDEX_EXTRA_TRANS_BLOCKS + - FSFILT_DATA_TRANS_BLOCKS(inode->i_sb); - break; default: CERROR("unknown transaction start op %d\n", op); LBUG(); } @@ -253,7 +336,7 @@ static void *fsfilt_ext3_start(struct inode *inode, int op, void *desc_private, journal_start: LASSERTF(nblocks > 0, "can't start %d credit transaction\n", nblocks); - handle = fsfilt_ext3_journal_start(inode, nblocks); + handle = ext3_journal_start(inode, nblocks); if (!IS_ERR(handle)) LASSERT(current->journal_info == handle); @@ -339,7 +422,7 @@ static int fsfilt_ext3_credits_needed(int objcount, struct fsfilt_objinfo *fso, /* We assume that there will be 1 bit set in s_dquot.flags for each * quota file that is active. This is at least true for now. */ - needed += hweight32(sb_any_quota_enabled(sb)) * + needed += hweight32(ll_sb_any_quota_active(sb)) * FSFILT_SINGLEDATA_TRANS_BLOCKS(sb); #endif @@ -390,7 +473,7 @@ static void *fsfilt_ext3_brw_start(int objcount, struct fsfilt_objinfo *fso, } LASSERTF(needed > 0, "can't start %d credit transaction\n", needed); - handle = fsfilt_ext3_journal_start(fso->fso_dentry->d_inode, needed); + handle = ext3_journal_start(fso->fso_dentry->d_inode, needed); if (IS_ERR(handle)) { CERROR("can't get handle for %d credits: rc = %ld\n", needed, PTR_ERR(handle)); @@ -414,11 +497,11 @@ static int fsfilt_ext3_extend(struct inode *inode, unsigned int nblocks,void *h) if (handle->h_buffer_credits > nblocks) return 0; - if (journal_extend(handle, nblocks) == 0) + if (ext3_journal_extend(handle, nblocks) == 0) return 0; ext3_mark_inode_dirty(handle, inode); - return journal_restart(handle, nblocks); + return ext3_journal_restart(handle, nblocks); } static int fsfilt_ext3_commit(struct inode *inode, void *h, int force_sync) @@ -430,7 +513,7 @@ static int fsfilt_ext3_commit(struct inode *inode, void *h, int force_sync) if (force_sync) handle->h_sync = 1; /* recovery likes this */ - rc = fsfilt_ext3_journal_stop(handle); + rc = ext3_journal_stop(handle); return rc; } @@ -451,12 +534,12 @@ static int fsfilt_ext3_commit_async(struct inode *inode, void *h, tid = transaction->t_tid; /* we don't want to be blocked */ handle->h_sync = 0; - rc = fsfilt_ext3_journal_stop(handle); + rc = ext3_journal_stop(handle); if (rc) { CERROR("error while stopping transaction: %d\n", rc); return rc; } - log_start_commit(journal, tid); + fsfilt_log_start_commit(journal, tid); *wait_handle = (void *) tid; CDEBUG(D_INODE, "commit async: %lu\n", (unsigned long) tid); @@ -472,7 +555,7 @@ static int fsfilt_ext3_commit_wait(struct inode *inode, void *h) if (unlikely(is_journal_aborted(journal))) return -EIO; - log_wait_commit(EXT3_JOURNAL(inode), tid); + fsfilt_log_wait_commit(EXT3_JOURNAL(inode), tid); if (unlikely(is_journal_aborted(journal))) return -EIO; @@ -508,7 +591,8 @@ static int fsfilt_ext3_setattr(struct dentry *dentry, void *handle, if (iattr->ia_valid & ATTR_MODE) { inode->i_mode = iattr->ia_mode; - if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID)) + if (!cfs_curproc_is_in_groups(inode->i_gid) && + !cfs_capable(CFS_CAP_FSETID)) inode->i_mode &= ~S_ISGID; } @@ -538,20 +622,38 @@ static int fsfilt_ext3_setattr(struct dentry *dentry, void *handle, RETURN(rc); } -static int fsfilt_ext3_iocontrol(struct inode * inode, struct file *file, +static int fsfilt_ext3_iocontrol(struct inode *inode, struct file *file, unsigned int cmd, unsigned long arg) { int rc = 0; ENTRY; /* FIXME: Can't do this because of nested transaction deadlock */ - if (cmd == EXT3_IOC_SETFLAGS && (*(int *)arg) & EXT3_JOURNAL_DATA_FL) { - CERROR("can't set data journal flag on file\n"); - RETURN(-EPERM); + if (cmd == EXT3_IOC_SETFLAGS) { + /* We can't enable data journaling on OST objects, because + * this forces the transaction to be closed in order to + * flush the journal, but the caller will already have a + * compound transaction open to update the last_rcvd file, + * and this thread would deadlock trying to set the flag. */ + if ((*(int *)arg) & EXT3_JOURNAL_DATA_FL) { + CERROR("can't set data journal flag on file\n"); + RETURN(-EPERM); + } + /* Because the MDS does not see the EXTENTS_FL set on the + * OST objects, mask this flag into all set flags. It is + * not legal to clear this flag in any case, so we are not + * changing the functionality by doing this. b=22911 */ + *(int *)arg |= EXT3_I(inode)->i_flags & EXT3_EXTENTS_FL; } +#ifdef HAVE_EXT4_LDISKFS + /* ext4_ioctl does not have a inode argument */ + if (inode->i_fop->unlocked_ioctl) + rc = inode->i_fop->unlocked_ioctl(file, cmd, arg); +#else if (inode->i_fop->ioctl) rc = inode->i_fop->ioctl(inode, file, cmd, arg); +#endif else RETURN(-ENOTTY); @@ -566,7 +668,7 @@ static int fsfilt_ext3_set_md(struct inode *inode, void *handle, LASSERT(TRYLOCK_INODE_MUTEX(inode) == 0); rc = ext3_xattr_set_handle(handle, inode, EXT3_XATTR_INDEX_TRUSTED, - name, lmm, lmm_size, 0); + name, lmm, lmm_size, XATTR_NO_CTIME); if (rc && rc != -EROFS) @@ -619,8 +721,8 @@ static ssize_t fsfilt_ext3_readpage(struct file *file, char *buf, size_t count, const int blkbits = inode->i_sb->s_blocksize_bits; const int blksize = inode->i_sb->s_blocksize; - CDEBUG(D_EXT2, "reading "LPSZ" at dir %lu+%llu\n", - count, inode->i_ino, *off); + CDEBUG(D_EXT2, "reading %lu at dir %lu+%llu\n", + (unsigned long)count, inode->i_ino, *off); while (count > 0) { struct buffer_head *bh; @@ -678,7 +780,7 @@ static int fsfilt_ext3_add_journal_cb(struct obd_device *obd, __u64 last_rcvd, { struct fsfilt_cb_data *fcb; - OBD_SLAB_ALLOC(fcb, fcb_cache, CFS_ALLOC_IO, sizeof *fcb); + OBD_SLAB_ALLOC_PTR_GFP(fcb, fcb_cache, CFS_ALLOC_IO); if (fcb == NULL) RETURN(-ENOMEM); @@ -688,8 +790,8 @@ static int fsfilt_ext3_add_journal_cb(struct obd_device *obd, __u64 last_rcvd, fcb->cb_data = cb_data; CDEBUG(D_EXT2, "set callback for last_rcvd: "LPD64"\n", last_rcvd); - journal_callback_set(handle, fsfilt_ext3_cb_func, - (struct journal_callback *)fcb); + fsfilt_journal_callback_set(handle, fsfilt_ext3_cb_func, + (struct journal_callback *)fcb); return 0; } @@ -707,9 +809,7 @@ static int fsfilt_ext3_statfs(struct super_block *sb, struct obd_statfs *osfs) int rc; memset(&sfs, 0, sizeof(sfs)); - rc = ll_do_statfs(sb, &sfs); - if (!rc && sfs.f_bfree < sfs.f_ffree) { sfs.f_files = (sfs.f_files - sfs.f_ffree) + sfs.f_bfree; sfs.f_ffree = sfs.f_bfree; @@ -724,21 +824,26 @@ static int fsfilt_ext3_sync(struct super_block *sb) return ext3_force_commit(sb); } -#if defined(EXT3_MULTIBLOCK_ALLOCATOR) && (!defined(EXT3_EXT_CACHE_NO) || defined(EXT_CACHE_MARK)) -#warning "kernel code has old extents/mballoc patch, disabling" -#undef EXT3_MULTIBLOCK_ALLOCATOR -#endif #ifndef EXT3_EXTENTS_FL #define EXT3_EXTENTS_FL 0x00080000 /* Inode uses extents */ #endif -#ifdef EXT3_MULTIBLOCK_ALLOCATOR #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,17)) -#define ext3_up_truncate_sem(inode) up(&EXT3_I(inode)->truncate_sem); -#define ext3_down_truncate_sem(inode) down(&EXT3_I(inode)->truncate_sem); +# define fsfilt_up_truncate_sem(inode) up(&LDISKFS_I(inode)->truncate_sem); +# define fsfilt_down_truncate_sem(inode) down(&LDISKFS_I(inode)->truncate_sem); #else -#define ext3_up_truncate_sem(inode) mutex_unlock(&EXT3_I(inode)->truncate_mutex); -#define ext3_down_truncate_sem(inode) mutex_lock(&EXT3_I(inode)->truncate_mutex); +# ifdef HAVE_EXT4_LDISKFS +# ifdef WALK_SPACE_HAS_DATA_SEM /* We only use it in fsfilt_map_nblocks() for now */ +# define fsfilt_up_truncate_sem(inode) do{ }while(0) +# define fsfilt_down_truncate_sem(inode) do{ }while(0) +# else +# define fsfilt_up_truncate_sem(inode) up_write((&EXT4_I(inode)->i_data_sem)) +# define fsfilt_down_truncate_sem(inode) down_write((&EXT4_I(inode)->i_data_sem)) +# endif +# else +# define fsfilt_up_truncate_sem(inode) mutex_unlock(&EXT3_I(inode)->truncate_mutex) +# define fsfilt_down_truncate_sem(inode) mutex_lock(&EXT3_I(inode)->truncate_mutex) +# endif #endif #ifndef EXT_ASSERT @@ -747,10 +852,14 @@ static int fsfilt_ext3_sync(struct super_block *sb) #ifdef EXT3_EXT_HAS_NO_TREE /* for kernels 2.6.18 and later */ +#ifdef HAVE_EXT4_LDISKFS +#define EXT_GENERATION(inode) (EXT4_I(inode)->i_ext_generation) +#else +#define EXT_GENERATION(inode) ext_generation(inode) +#endif #define ext3_ext_base inode #define ext3_ext_base2inode(inode) (inode) #define EXT_DEPTH(inode) ext_depth(inode) -#define EXT_GENERATION(inode) ext_generation(inode) #define fsfilt_ext3_ext_walk_space(inode, block, num, cb, cbdata) \ ext3_ext_walk_space(inode, block, num, cb, cbdata); #else @@ -761,15 +870,6 @@ static int fsfilt_ext3_sync(struct super_block *sb) #endif #include -#if EXT3_EXT_MAGIC == 0xf301 -#define ee_start e_start -#define ee_block e_block -#define ee_len e_num -#endif -#ifndef EXT3_BB_MAX_BLOCKS -#define ext3_mb_new_blocks(handle, inode, goal, count, aflags, err) \ - ext3_new_blocks(handle, inode, count, goal, err) -#endif struct bpointers { unsigned long *blocks; @@ -780,8 +880,8 @@ struct bpointers { int create; }; -static int ext3_ext_find_goal(struct inode *inode, struct ext3_ext_path *path, - unsigned long block, int *aflags) +static long ext3_ext_find_goal(struct inode *inode, struct ext3_ext_path *path, + unsigned long block, int *aflags) { struct ext3_inode_info *ei = EXT3_I(inode); unsigned long bg_start; @@ -793,18 +893,8 @@ static int ext3_ext_find_goal(struct inode *inode, struct ext3_ext_path *path, depth = path->p_depth; /* try to predict block placement */ - if ((ex = path[depth].p_ext)) { -#if 0 - /* This prefers to eat into a contiguous extent - * rather than find an extent that the whole - * request will fit into. This can fragment data - * block allocation and prevents our lovely 1M I/Os - * from reaching the disk intact. */ - if (ex->ee_block + ex->ee_len == block) - *aflags |= 1; -#endif - return ex->ee_start + (block - ex->ee_block); - } + if ((ex = path[depth].p_ext)) + return ext_pblock(ex) + (block - le32_to_cpu(ex->ee_block)); /* it looks index is empty * try to find starting from index itself */ @@ -867,7 +957,6 @@ static unsigned long new_blocks(handle_t *handle, struct ext3_ext_base *base, pblock = ext3_mb_new_blocks(handle, &ar, err); *count = ar.len; return pblock; - } #endif @@ -875,13 +964,20 @@ static unsigned long new_blocks(handle_t *handle, struct ext3_ext_base *base, static int ext3_ext_new_extent_cb(struct ext3_ext_base *base, struct ext3_ext_path *path, struct ext3_ext_cache *cex, +#ifdef HAVE_EXT_PREPARE_CB_EXTENT + struct ext3_extent *ex, +#endif void *cbdata) { struct bpointers *bp = cbdata; #else static int ext3_ext_new_extent_cb(struct ext3_ext_base *base, struct ext3_ext_path *path, - struct ext3_ext_cache *cex) + struct ext3_ext_cache *cex +#ifdef HAVE_EXT_PREPARE_CB_EXTENT + , struct ext3_extent *ex +#endif + ) { struct bpointers *bp = base->private; #endif @@ -923,18 +1019,18 @@ static int ext3_ext_new_extent_cb(struct ext3_ext_base *base, tgen = EXT_GENERATION(base); count = ext3_ext_calc_credits_for_insert(base, path); - ext3_up_truncate_sem(inode); + fsfilt_up_truncate_sem(inode); - handle = fsfilt_ext3_journal_start(inode, count+EXT3_ALLOC_NEEDED+1); + handle = ext3_journal_start(inode, count+EXT3_ALLOC_NEEDED+1); if (IS_ERR(handle)) { - ext3_down_truncate_sem(inode); + fsfilt_down_truncate_sem(inode); return PTR_ERR(handle); } - ext3_down_truncate_sem(inode); + fsfilt_down_truncate_sem(inode); if (tgen != EXT_GENERATION(base)) { /* the tree has changed. so path can be invalid at moment */ - fsfilt_ext3_journal_stop(handle); + ext3_journal_stop(handle); return EXT_REPEAT; } @@ -945,14 +1041,19 @@ static int ext3_ext_new_extent_cb(struct ext3_ext_base *base, EXT_ASSERT(count <= cex->ec_len); /* insert new extent */ - nex.ee_block = cex->ec_block; - nex.ee_start = pblock; - nex.ee_len = count; - err = ext3_ext_insert_extent(handle, base, path, &nex); + nex.ee_block = cpu_to_le32(cex->ec_block); + ext3_ext_store_pblock(&nex, pblock); + nex.ee_len = cpu_to_le16(count); + err = fsfilt_ext3_ext_insert_extent(handle, base, path, &nex, 0); if (err) { - CERROR("can't insert extent: %d\n", err); - /* XXX: export ext3_free_blocks() */ - /*ext3_free_blocks(handle, inode, nex.ee_start, nex.ee_len, 0);*/ + /* free data blocks we just allocated */ + /* not a good idea to call discard here directly, + * but otherwise we'd need to call it every free() */ +#ifdef EXT3_MB_HINT_GROUP_ALLOC + ext3_mb_discard_inode_preallocations(inode); +#endif + ext3_free_blocks(handle, inode, ext_pblock(&nex), + cpu_to_le16(nex.ee_len), 0); goto out; } @@ -961,13 +1062,13 @@ static int ext3_ext_new_extent_cb(struct ext3_ext_base *base, * we are asking ext3_ext_walk_space() to continue * scaning after that block */ - cex->ec_len = nex.ee_len; - cex->ec_start = nex.ee_start; - BUG_ON(nex.ee_len == 0); - BUG_ON(nex.ee_block != cex->ec_block); + cex->ec_len = le16_to_cpu(nex.ee_len); + cex->ec_start = ext_pblock(&nex); + BUG_ON(le16_to_cpu(nex.ee_len) == 0); + BUG_ON(le32_to_cpu(nex.ee_block) != cex->ec_block); out: - fsfilt_ext3_journal_stop(handle); + ext3_journal_stop(handle); map: if (err >= 0) { /* map blocks */ @@ -975,9 +1076,10 @@ map: CERROR("hmm. why do we find this extent?\n"); CERROR("initial space: %lu:%u\n", bp->start, bp->init_num); - CERROR("current extent: %u/%u/%u %d\n", + CERROR("current extent: %u/%u/%llu %d\n", cex->ec_block, cex->ec_len, - cex->ec_start, cex->ec_type); + (unsigned long long)cex->ec_start, + cex->ec_type); } i = 0; if (cex->ec_block < bp->start) @@ -1031,11 +1133,11 @@ int fsfilt_map_nblocks(struct inode *inode, unsigned long block, bp.init_num = bp.num = num; bp.create = create; - ext3_down_truncate_sem(inode); + fsfilt_down_truncate_sem(inode); err = fsfilt_ext3_ext_walk_space(base, block, num, ext3_ext_new_extent_cb, &bp); ext3_ext_invalidate_cache(base); - ext3_up_truncate_sem(inode); + fsfilt_up_truncate_sem(inode); return err; } @@ -1089,7 +1191,6 @@ int fsfilt_ext3_map_ext_inode_pages(struct inode *inode, struct page **page, cleanup: return rc; } -#endif /* EXT3_MULTIBLOCK_ALLOCATOR */ extern int ext3_map_inode_page(struct inode *inode, struct page *page, unsigned long *blocks, int *created, int create); @@ -1118,22 +1219,21 @@ int fsfilt_ext3_map_bm_inode_pages(struct inode *inode, struct page **page, int fsfilt_ext3_map_inode_pages(struct inode *inode, struct page **page, int pages, unsigned long *blocks, int *created, int create, - struct semaphore *optional_sem) + cfs_semaphore_t *optional_sem) { int rc; -#ifdef EXT3_MULTIBLOCK_ALLOCATOR + if (EXT3_I(inode)->i_flags & EXT3_EXTENTS_FL) { rc = fsfilt_ext3_map_ext_inode_pages(inode, page, pages, blocks, created, create); return rc; } -#endif if (optional_sem != NULL) - down(optional_sem); + cfs_down(optional_sem); rc = fsfilt_ext3_map_bm_inode_pages(inode, page, pages, blocks, created, create); if (optional_sem != NULL) - up(optional_sem); + cfs_up(optional_sem); return rc; } @@ -1145,19 +1245,19 @@ int fsfilt_ext3_read(struct inode *inode, void *buf, int size, loff_t *offs) int err, blocksize, csize, boffs, osize = size; /* prevent reading after eof */ - lock_kernel(); + cfs_lock_kernel(); if (i_size_read(inode) < *offs + size) { size = i_size_read(inode) - *offs; - unlock_kernel(); + cfs_unlock_kernel(); if (size < 0) { - CERROR("size %llu is too short for read %u@%llu\n", - i_size_read(inode), size, *offs); - return -EIO; + CDEBUG(D_EXT2, "size %llu is too short for read @%llu\n", + i_size_read(inode), *offs); + return -EBADR; } else if (size == 0) { return 0; } } else { - unlock_kernel(); + cfs_unlock_kernel(); } blocksize = 1 << inode->i_blkbits; @@ -1240,14 +1340,14 @@ int fsfilt_ext3_write_handle(struct inode *inode, void *buf, int bufsize, /* correct in-core and on-disk sizes */ if (new_size > i_size_read(inode)) { - lock_kernel(); + cfs_lock_kernel(); if (new_size > i_size_read(inode)) i_size_write(inode, new_size); if (i_size_read(inode) > EXT3_I(inode)->i_disksize) EXT3_I(inode)->i_disksize = i_size_read(inode); if (i_size_read(inode) > old_size) mark_inode_dirty(inode); - unlock_kernel(); + cfs_unlock_kernel(); } if (err == 0) @@ -1268,7 +1368,7 @@ static int fsfilt_ext3_write_record(struct file *file, void *buf, int bufsize, block_count = (*offs & (blocksize - 1)) + bufsize; block_count = (block_count + blocksize - 1) >> inode->i_blkbits; - handle = fsfilt_ext3_journal_start(inode, + handle = ext3_journal_start(inode, block_count * FSFILT_DATA_TRANS_BLOCKS(inode->i_sb) + 2); if (IS_ERR(handle)) { CERROR("can't start transaction for %d blocks (%d bytes)\n", @@ -1281,26 +1381,44 @@ static int fsfilt_ext3_write_record(struct file *file, void *buf, int bufsize, if (!err && force_sync) handle->h_sync = 1; /* recovery likes this */ - fsfilt_ext3_journal_stop(handle); + ext3_journal_stop(handle); return err; } static int fsfilt_ext3_setup(struct super_block *sb) { +#if ((LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,6)) && \ + defined(HAVE_QUOTA_SUPPORT)) || defined(S_PDIROPS) + struct ext3_sb_info *sbi = EXT3_SB(sb); #if 0 - EXT3_SB(sb)->dx_lock = fsfilt_ext3_dx_lock; - EXT3_SB(sb)->dx_unlock = fsfilt_ext3_dx_unlock; + sbi->dx_lock = fsfilt_ext3_dx_lock; + sbi->dx_unlock = fsfilt_ext3_dx_unlock; +#endif #endif #ifdef S_PDIROPS CWARN("Enabling PDIROPS\n"); - set_opt(EXT3_SB(sb)->s_mount_opt, PDIROPS); + set_opt(sbi->s_mount_opt, PDIROPS); sb->s_flags |= S_PDIROPS; #endif if (!EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_DIR_INDEX)) CWARN("filesystem doesn't have dir_index feature enabled\n"); -#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,13)) && HAVE_QUOTA_SUPPORT - set_opt(EXT3_SB(sb)->s_mount_opt, QUOTA); +#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,6)) && defined(HAVE_QUOTA_SUPPORT) + /* enable journaled quota support */ + /* kfreed in ext3_put_super() */ + sbi->s_qf_names[USRQUOTA] = kstrdup("lquota.user.reserved", GFP_KERNEL); + if (!sbi->s_qf_names[USRQUOTA]) + return -ENOMEM; + sbi->s_qf_names[GRPQUOTA] = kstrdup("lquota.group.reserved", GFP_KERNEL); + if (!sbi->s_qf_names[GRPQUOTA]) { + kfree(sbi->s_qf_names[USRQUOTA]); + sbi->s_qf_names[USRQUOTA] = NULL; + return -ENOMEM; + } + sbi->s_jquota_fmt = QFMT_VFS_V0; +#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,13)) + set_opt(sbi->s_mount_opt, QUOTA); +#endif #endif return 0; } @@ -1336,8 +1454,7 @@ static int fsfilt_ext3_get_op_len(int op, struct fsfilt_objinfo *fso, int logs) return 0; } -static const char *op_quotafile[] = { "lquota.user", "lquota.group" }; - +#ifdef HAVE_QUOTA_SUPPORT #define DQINFO_COPY(out, in) \ do { \ Q_COPY(out, in, dqi_bgrace); \ @@ -1359,8 +1476,6 @@ do { \ Q_COPY(out, in, dqb_valid); \ } while (0) - - static int fsfilt_ext3_quotactl(struct super_block *sb, struct obd_quotactl *oqc) { @@ -1392,14 +1507,14 @@ static int fsfilt_ext3_quotactl(struct super_block *sb, continue; if (oqc->qc_cmd == Q_QUOTAON) { - if (!qcop->quota_on) - GOTO(out, rc = -ENOSYS); - rc = qcop->quota_on(sb, i, oqc->qc_id, - (char *)op_quotafile[i]); + char *name[MAXQUOTAS] = LUSTRE_OPQFILES_NAMES_V2; + + LASSERT(oqc->qc_id == LUSTRE_QUOTA_V2); + + rc = ll_quota_on(sb, i, QFMT_VFS_V0, + name[i], 0); } else if (oqc->qc_cmd == Q_QUOTAOFF) { - if (!qcop->quota_off) - GOTO(out, rc = -ENOSYS); - rc = qcop->quota_off(sb, i); + rc = ll_quota_off(sb, i, 0); } if (rc == -EBUSY) @@ -1428,14 +1543,38 @@ static int fsfilt_ext3_quotactl(struct super_block *sb, if (!qcop->get_dqblk) GOTO(out, rc = -ENOSYS); rc = qcop->get_dqblk(sb, oqc->qc_type, oqc->qc_id, dqblk); + if (!rc) + dqblk->dqb_valid = QIF_LIMITS | QIF_USAGE; break; case Q_SYNC: if (!sb->s_qcop->quota_sync) GOTO(out, rc = -ENOSYS); qcop->quota_sync(sb, oqc->qc_type); break; + case Q_FINVALIDATE: + CDEBUG(D_WARNING, "invalidating operational quota files\n"); + for (i = 0; i < MAXQUOTAS; i++) { + struct file *fp; + char *name[MAXQUOTAS] = LUSTRE_OPQFILES_NAMES_V2; + + LASSERT(oqc->qc_id == LUSTRE_QUOTA_V2); + + if (!Q_TYPESET(oqc, i)) + continue; + + fp = filp_open(name[i], O_CREAT | O_TRUNC | O_RDWR, 0644); + if (IS_ERR(fp)) { + rc = PTR_ERR(fp); + CERROR("error invalidating operational quota file" + " %s (rc:%d)\n", name[i], rc); + } else { + filp_close(fp, 0); + } + + } + break; default: - CERROR("unsupported quotactl command: %d", oqc->qc_cmd); + CERROR("unsupported quotactl command: %d\n", oqc->qc_cmd); LBUG(); } out: @@ -1446,26 +1585,26 @@ out: OBD_FREE_PTR(dqblk); if (rc) - CDEBUG(D_QUOTA, "quotactl command %#x, id %u, type %d " + CDEBUG(D_QUOTA, "quotactl command %#x, id %u, type %u " "failed: %d\n", oqc->qc_cmd, oqc->qc_id, oqc->qc_type, rc); RETURN(rc); } struct chk_dqblk{ - struct hlist_node dqb_hash; /* quotacheck hash */ - struct list_head dqb_list; /* in list also */ - qid_t dqb_id; /* uid/gid */ - short dqb_type; /* USRQUOTA/GRPQUOTA */ - __u32 dqb_bhardlimit; /* block hard limit */ - __u32 dqb_bsoftlimit; /* block soft limit */ - qsize_t dqb_curspace; /* current space */ - __u32 dqb_ihardlimit; /* inode hard limit */ - __u32 dqb_isoftlimit; /* inode soft limit */ - __u32 dqb_curinodes; /* current inodes */ - __u64 dqb_btime; /* block grace time */ - __u64 dqb_itime; /* inode grace time */ - __u32 dqb_valid; /* flag for above fields */ + cfs_hlist_node_t dqb_hash; /** quotacheck hash */ + cfs_list_t dqb_list; /** in list also */ + qid_t dqb_id; /** uid/gid */ + short dqb_type; /** USRQUOTA/GRPQUOTA */ + qsize_t dqb_bhardlimit; /** block hard limit */ + qsize_t dqb_bsoftlimit; /** block soft limit */ + qsize_t dqb_curspace; /** current space */ + qsize_t dqb_ihardlimit; /** inode hard limit */ + qsize_t dqb_isoftlimit; /** inode soft limit */ + qsize_t dqb_curinodes; /** current inodes */ + __u64 dqb_btime; /** block grace time */ + __u64 dqb_itime; /** inode grace time */ + __u32 dqb_valid; /** flag for above fields */ }; static inline unsigned int chkquot_hash(qid_t id, int type) @@ -1477,13 +1616,13 @@ static inline unsigned int chkquot_hash(qid_t id, int type) } static inline struct chk_dqblk * -find_chkquot(struct hlist_head *head, qid_t id, int type) +find_chkquot(cfs_hlist_head_t *head, qid_t id, int type) { - struct hlist_node *node; + cfs_hlist_node_t *node; struct chk_dqblk *cdqb; - hlist_for_each(node, head) { - cdqb = hlist_entry(node, struct chk_dqblk, dqb_hash); + cfs_hlist_for_each(node, head) { + cdqb = cfs_hlist_entry(node, struct chk_dqblk, dqb_hash); if (cdqb->dqb_id == id && cdqb->dqb_type == type) return cdqb; } @@ -1497,8 +1636,8 @@ static struct chk_dqblk *alloc_chkquot(qid_t id, int type) OBD_ALLOC_PTR(cdqb); if (cdqb) { - INIT_HLIST_NODE(&cdqb->dqb_hash); - INIT_LIST_HEAD(&cdqb->dqb_list); + CFS_INIT_HLIST_NODE(&cdqb->dqb_hash); + CFS_INIT_LIST_HEAD(&cdqb->dqb_list); cdqb->dqb_id = id; cdqb->dqb_type = type; } @@ -1507,10 +1646,10 @@ static struct chk_dqblk *alloc_chkquot(qid_t id, int type) } static struct chk_dqblk * -cqget(struct super_block *sb, struct hlist_head *hash, struct list_head *list, - qid_t id, int type, int first_check) +cqget(struct super_block *sb, cfs_hlist_head_t *hash, + cfs_list_t *list, qid_t id, int type, int first_check) { - struct hlist_head *head = hash + chkquot_hash(id, type); + cfs_hlist_head_t *head = hash + chkquot_hash(id, type); struct if_dqblk dqb; struct chk_dqblk *cdqb; int rc; @@ -1535,13 +1674,13 @@ cqget(struct super_block *sb, struct hlist_head *hash, struct list_head *list, } } - hlist_add_head(&cdqb->dqb_hash, head); - list_add_tail(&cdqb->dqb_list, list); + cfs_hlist_add_head(&cdqb->dqb_hash, head); + cfs_list_add_tail(&cdqb->dqb_list, list); return cdqb; } -static inline int quota_onoff(struct super_block *sb, int cmd, int type) +static inline int quota_onoff(struct super_block *sb, int cmd, int type, int qfmt) { struct obd_quotactl *oqctl; int rc; @@ -1551,7 +1690,7 @@ static inline int quota_onoff(struct super_block *sb, int cmd, int type) RETURN(-ENOMEM); oqctl->qc_cmd = cmd; - oqctl->qc_id = QFMT_LDISKFS; + oqctl->qc_id = qfmt; oqctl->qc_type = type; rc = fsfilt_ext3_quotactl(sb, oqctl); @@ -1580,8 +1719,9 @@ static inline int read_old_dqinfo(struct super_block *sb, int type, RETURN(rc); } +#ifndef HAVE_EXT4_LDISKFS static inline struct ext3_group_desc * -get_group_desc(struct super_block *sb, int group) +get_group_desc(struct super_block *sb, int group, struct buffer_head **bh) { unsigned long desc_block, desc; struct ext3_group_desc *gdp; @@ -1595,34 +1735,29 @@ get_group_desc(struct super_block *sb, int group) } static inline struct buffer_head * -read_inode_bitmap(struct super_block *sb, unsigned long group) +ext3_read_inode_bitmap(struct super_block *sb, unsigned long group) { struct ext3_group_desc *desc; struct buffer_head *bh; - desc = get_group_desc(sb, group); - bh = sb_bread(sb, le32_to_cpu(desc->bg_inode_bitmap)); - + desc = get_group_desc(sb, group, NULL); + bh = sb_bread(sb, ext3_inode_bitmap(sb, desc)); return bh; } -static inline struct inode *ext3_iget_inuse(struct super_block *sb, - struct buffer_head *bitmap_bh, - int index, unsigned long ino) -{ - struct inode *inode = NULL; - - if (ext3_test_bit(index, bitmap_bh->b_data)) - inode = iget(sb, ino); - - return inode; +static __u32 ext3_itable_unused_count(struct super_block *sb, + struct ext3_group_desc *bg) { + return le16_to_cpu(bg->bg_itable_unused); } +#else +#define get_group_desc ext3_get_group_desc +#endif struct qchk_ctxt { - struct hlist_head qckt_hash[NR_DQHASH]; /* quotacheck hash */ - struct list_head qckt_list; /* quotacheck list */ + cfs_hlist_head_t qckt_hash[NR_DQHASH]; /* quotacheck hash */ + cfs_list_t qckt_list; /* quotacheck list */ int qckt_first_check[MAXQUOTAS]; /* 1 if no old quotafile */ - struct if_dqinfo qckt_dqinfo[MAXQUOTAS]; /* old dqinfo */ + struct if_dqinfo qckt_dqinfo[MAXQUOTAS]; /* old dqinfo */ }; static int add_inode_quota(struct inode *inode, struct qchk_ctxt *qctxt, @@ -1673,24 +1808,8 @@ static int add_inode_quota(struct inode *inode, struct qchk_ctxt *qctxt, return rc; } -static int v2_write_dqheader(struct file *f, int type) -{ - static const __u32 quota_magics[] = V2_INITQMAGICS; - static const __u32 quota_versions[] = V2_INITQVERSIONS; - struct v2_disk_dqheader dqhead; - loff_t offset = 0; - - CLASSERT(ARRAY_SIZE(quota_magics) == ARRAY_SIZE(quota_versions)); - LASSERT(0 <= type && type < ARRAY_SIZE(quota_magics)); - - dqhead.dqh_magic = cpu_to_le32(quota_magics[type]); - dqhead.dqh_version = cpu_to_le32(quota_versions[type]); - - return cfs_user_write(f, (char *)&dqhead, sizeof(dqhead), &offset); -} - /* write dqinfo struct in a new quota file */ -static int v2_write_dqinfo(struct file *f, int type, struct if_dqinfo *info) +static int v3_write_dqinfo(struct file *f, int type, struct if_dqinfo *info) { struct v2_disk_dqinfo dqinfo; __u32 blocks = V2_DQTREEOFF + 1; @@ -1714,6 +1833,22 @@ static int v2_write_dqinfo(struct file *f, int type, struct if_dqinfo *info) return cfs_user_write(f, (char *)&dqinfo, sizeof(dqinfo), &offset); } +static int v3_write_dqheader(struct file *f, int type) +{ + static const __u32 quota_magics[] = V2_INITQMAGICS; + static const __u32 quota_versions[] = V2_INITQVERSIONS_R1; + struct v2_disk_dqheader dqhead; + loff_t offset = 0; + + CLASSERT(ARRAY_SIZE(quota_magics) == ARRAY_SIZE(quota_versions)); + LASSERT(0 <= type && type < ARRAY_SIZE(quota_magics)); + + dqhead.dqh_magic = cpu_to_le32(quota_magics[type]); + dqhead.dqh_version = cpu_to_le32(quota_versions[type]); + + return cfs_user_write(f, (char *)&dqhead, sizeof(dqhead), &offset); +} + static int create_new_quota_files(struct qchk_ctxt *qctxt, struct obd_quotactl *oqc) { @@ -1724,32 +1859,36 @@ static int create_new_quota_files(struct qchk_ctxt *qctxt, struct if_dqinfo *info = qctxt->qckt_first_check[i]? NULL : &qctxt->qckt_dqinfo[i]; struct file *file; + const char *name[MAXQUOTAS] = LUSTRE_OPQFILES_NAMES_V2; if (!Q_TYPESET(oqc, i)) continue; - file = filp_open(op_quotafile[i], O_RDWR | O_CREAT | O_TRUNC, - 0644); + LASSERT(oqc->qc_id == LUSTRE_QUOTA_V2); + + file = filp_open(name[i], O_RDWR | O_CREAT | O_TRUNC, 0644); if (IS_ERR(file)) { rc = PTR_ERR(file); CERROR("can't create %s file: rc = %d\n", - op_quotafile[i], rc); + name[i], rc); GOTO(out, rc); } if (!S_ISREG(file->f_dentry->d_inode->i_mode)) { - CERROR("file %s is not regular", op_quotafile[i]); + CERROR("file %s is not regular", name[i]); filp_close(file, 0); GOTO(out, rc = -EINVAL); } - rc = v2_write_dqheader(file, i); + DQUOT_DROP(file->f_dentry->d_inode); + + rc = v3_write_dqheader(file, i); if (rc) { filp_close(file, 0); GOTO(out, rc); } - rc = v2_write_dqinfo(file, i, info); + rc = v3_write_dqinfo(file, i, info); filp_close(file, 0); if (rc) GOTO(out, rc); @@ -1777,13 +1916,13 @@ static int commit_chkquot(struct super_block *sb, struct qchk_ctxt *qctxt, if (cdqb->dqb_bsoftlimit && toqb(cdqb->dqb_curspace) >= cdqb->dqb_bsoftlimit && !cdqb->dqb_btime) - cdqb->dqb_btime = + cdqb->dqb_btime = now + qctxt->qckt_dqinfo[cdqb->dqb_type].dqi_bgrace; if (cdqb->dqb_isoftlimit && cdqb->dqb_curinodes >= cdqb->dqb_isoftlimit && !cdqb->dqb_itime) - cdqb->dqb_itime = + cdqb->dqb_itime = now + qctxt->qckt_dqinfo[cdqb->dqb_type].dqi_igrace; cdqb->dqb_valid = QIF_ALL; @@ -1804,28 +1943,32 @@ static int prune_chkquots(struct super_block *sb, struct chk_dqblk *cdqb, *tmp; int rc; - list_for_each_entry_safe(cdqb, tmp, &qctxt->qckt_list, dqb_list) { + cfs_list_for_each_entry_safe(cdqb, tmp, &qctxt->qckt_list, dqb_list) { if (!error) { rc = commit_chkquot(sb, qctxt, cdqb); if (rc) error = rc; } - hlist_del_init(&cdqb->dqb_hash); - list_del(&cdqb->dqb_list); + cfs_hlist_del_init(&cdqb->dqb_hash); + cfs_list_del(&cdqb->dqb_list); OBD_FREE_PTR(cdqb); } return error; } +#ifndef EXT3_FEATURE_RO_COMPAT_GDT_CSUM +#define EXT3_FEATURE_RO_COMPAT_GDT_CSUM 0x0010 +#endif + static int fsfilt_ext3_quotacheck(struct super_block *sb, struct obd_quotactl *oqc) { struct ext3_sb_info *sbi = EXT3_SB(sb); - int i, group; + int i, group, uninit_feat = 0; struct qchk_ctxt *qctxt; struct buffer_head *bitmap_bh = NULL; - unsigned long ino; + unsigned long ino, inode_inuse; struct inode *inode; int rc = 0; ENTRY; @@ -1838,39 +1981,81 @@ static int fsfilt_ext3_quotacheck(struct super_block *sb, } for (i = 0; i < NR_DQHASH; i++) - INIT_HLIST_HEAD(&qctxt->qckt_hash[i]); - INIT_LIST_HEAD(&qctxt->qckt_list); + CFS_INIT_HLIST_HEAD(&qctxt->qckt_hash[i]); + CFS_INIT_LIST_HEAD(&qctxt->qckt_list); for (i = 0; i < MAXQUOTAS; i++) { if (!Q_TYPESET(oqc, i)) continue; - rc = quota_onoff(sb, Q_QUOTAON, i); + rc = quota_onoff(sb, Q_QUOTAON, i, oqc->qc_id); if (!rc || rc == -EBUSY) { rc = read_old_dqinfo(sb, i, qctxt->qckt_dqinfo); if (rc) GOTO(out, rc); - } else if (rc == -ENOENT) { + } else if (rc == -ENOENT || rc == -EINVAL || rc == -EEXIST) { qctxt->qckt_first_check[i] = 1; } else if (rc) { GOTO(out, rc); } } + if (EXT3_HAS_RO_COMPAT_FEATURE(sb, EXT3_FEATURE_RO_COMPAT_GDT_CSUM)) + /* This filesystem supports the uninit group feature */ + uninit_feat = 1; + + /* number of inodes that have been allocated */ + inode_inuse = sbi->s_inodes_per_group * sbi->s_groups_count - + percpu_counter_sum(&sbi->s_freeinodes_counter); /* check quota and update in hash */ - for (group = 0; group < sbi->s_groups_count; group++) { + for (group = 0; group < sbi->s_groups_count && inode_inuse > 0; + group++) { + unsigned long used_count = sbi->s_inodes_per_group; + + if (uninit_feat) { + struct ext3_group_desc *desc; + desc = get_group_desc(sb, group, NULL); + if (!desc) + GOTO(out, -EIO); + + /* we don't really need to take the group lock here, + * but it may be useful if one day we support online + * quotacheck */ + spin_lock(sb_bgl_lock(sbi, group)); + if (desc->bg_flags & cpu_to_le16(EXT3_BG_INODE_UNINIT)) { + /* no inode in use in this group, just skip it */ + spin_unlock(sb_bgl_lock(sbi, group)); + continue; + } + used_count -= ext3_itable_unused_count(sb, desc); + spin_unlock(sb_bgl_lock(sbi, group)); + } + ino = group * sbi->s_inodes_per_group + 1; - bitmap_bh = read_inode_bitmap(sb, group); + bitmap_bh = ext3_read_inode_bitmap(sb, group); if (!bitmap_bh) { - CERROR("read_inode_bitmap group %d failed", group); - GOTO(out, rc = -EIO); + CERROR("%s: ext3_read_inode_bitmap group %d failed\n", + sb->s_id, group); + GOTO(out, -EIO); } - for (i = 0; i < sbi->s_inodes_per_group; i++, ino++) { + i = 0; + while (i < used_count && + (i = ext3_find_next_bit(bitmap_bh->b_data, + used_count, i)) < used_count) { + inode_inuse--; + i++; + ino = i + group * sbi->s_inodes_per_group; if (ino < sbi->s_first_ino) continue; +#if defined(HAVE_EXT4_LDISKFS) || !defined(HAVE_READ_INODE_IN_SBOPS) + inode = ext3_iget(sb, ino); +#else + inode = iget(sb, ino); +#endif + if (!inode || IS_ERR(inode)) + continue; - inode = ext3_iget_inuse(sb, bitmap_bh, i, ino); rc = add_inode_quota(inode, qctxt, oqc); iput(inode); if (rc) { @@ -1878,7 +2063,6 @@ static int fsfilt_ext3_quotacheck(struct super_block *sb, GOTO(out, rc); } } - brelse(bitmap_bh); } @@ -1886,7 +2070,7 @@ static int fsfilt_ext3_quotacheck(struct super_block *sb, * has limits but hasn't file) */ #ifdef HAVE_QUOTA_SUPPORT for (i = 0; i < MAXQUOTAS; i++) { - struct list_head id_list; + cfs_list_t id_list; struct dquot_id *dqid, *tmp; if (!Q_TYPESET(oqc, i)) @@ -1897,8 +2081,8 @@ static int fsfilt_ext3_quotacheck(struct super_block *sb, LASSERT(sb_dqopt(sb)->files[i] != NULL); - INIT_LIST_HEAD(&id_list); -#ifndef KERNEL_SUPPORTS_QUOTA_READ + CFS_INIT_LIST_HEAD(&id_list); +#ifndef KERNEL_SUPPORTS_QUOTA_READ rc = lustre_get_qids(sb_dqopt(sb)->files[i], NULL, i, &id_list); #else rc = lustre_get_qids(NULL, sb_dqopt(sb)->files[i], i, &id_list); @@ -1906,26 +2090,26 @@ static int fsfilt_ext3_quotacheck(struct super_block *sb, if (rc) CERROR("read old limits failed. (rc:%d)\n", rc); - list_for_each_entry_safe(dqid, tmp, &id_list, di_link) { - list_del_init(&dqid->di_link); + cfs_list_for_each_entry_safe(dqid, tmp, &id_list, di_link) { + cfs_list_del_init(&dqid->di_link); if (!rc) cqget(sb, qctxt->qckt_hash, &qctxt->qckt_list, dqid->di_id, i, qctxt->qckt_first_check[i]); - kfree(dqid); + OBD_FREE_PTR(dqid); } } #endif /* turn off quota cause we are to dump chk_dqblk to files */ - quota_onoff(sb, Q_QUOTAOFF, oqc->qc_type); + quota_onoff(sb, Q_QUOTAOFF, oqc->qc_type, oqc->qc_id); rc = create_new_quota_files(qctxt, oqc); if (rc) GOTO(out, rc); /* we use vfs functions to set dqblk, so turn quota on */ - rc = quota_onoff(sb, Q_QUOTAON, oqc->qc_type); + rc = quota_onoff(sb, Q_QUOTAON, oqc->qc_type, oqc->qc_id); out: /* dump and free chk_dqblk */ rc = prune_chkquots(sb, qctxt, rc); @@ -1933,7 +2117,7 @@ out: /* turn off quota, `lfs quotacheck` will turn on when all * nodes quotacheck finish. */ - quota_onoff(sb, Q_QUOTAOFF, oqc->qc_type); + quota_onoff(sb, Q_QUOTAOFF, oqc->qc_type, oqc->qc_id); oqc->qc_stat = rc; if (rc) @@ -1942,7 +2126,6 @@ out: RETURN(rc); } -#ifdef HAVE_QUOTA_SUPPORT static int fsfilt_ext3_quotainfo(struct lustre_quota_info *lqi, int type, int cmd) { @@ -1951,7 +2134,7 @@ static int fsfilt_ext3_quotainfo(struct lustre_quota_info *lqi, int type, if (lqi->qi_files[type] == NULL) { CERROR("operate qinfo before it's enabled!\n"); - RETURN(-EIO); + RETURN(-ESRCH); } switch (cmd) { @@ -1967,16 +2150,22 @@ static int fsfilt_ext3_quotainfo(struct lustre_quota_info *lqi, int type, case QFILE_INIT_INFO: rc = lustre_init_quota_info(lqi, type); break; + case QFILE_CONVERT: + rc = -ENOTSUPP; + CERROR("quota CONVERT command is not supported\n"); + break; default: - CERROR("Unsupported admin quota file cmd %d\n", cmd); - LBUG(); + rc = -ENOTSUPP; + CERROR("Unsupported admin quota file cmd %d\n" + "Are lquota.ko and fsfilt_ldiskfs.ko modules in sync?\n", + cmd); break; } RETURN(rc); } static int fsfilt_ext3_qids(struct file *file, struct inode *inode, int type, - struct list_head *list) + cfs_list_t *list) { return lustre_get_qids(file, inode, type, list); } @@ -1988,7 +2177,7 @@ static int fsfilt_ext3_dquot(struct lustre_dquot *dquot, int cmd) if (dquot->dq_info->qi_files[dquot->dq_type] == NULL) { CERROR("operate dquot before it's enabled!\n"); - RETURN(-EIO); + RETURN(-ESRCH); } switch (cmd) { @@ -2000,9 +2189,9 @@ static int fsfilt_ext3_dquot(struct lustre_dquot *dquot, int cmd) dquot->dq_dqb.dqb_isoftlimit || dquot->dq_dqb.dqb_bhardlimit || dquot->dq_dqb.dqb_bsoftlimit) - clear_bit(DQ_FAKE_B, &dquot->dq_flags); + cfs_clear_bit(DQ_FAKE_B, &dquot->dq_flags); else - set_bit(DQ_FAKE_B, &dquot->dq_flags); + cfs_set_bit(DQ_FAKE_B, &dquot->dq_flags); rc = lustre_commit_dquot(dquot); if (rc >= 0) @@ -2015,6 +2204,24 @@ static int fsfilt_ext3_dquot(struct lustre_dquot *dquot, int cmd) } RETURN(rc); } + +static int fsfilt_ext3_get_mblk(struct super_block *sb, int *count, + struct inode *inode, int frags) +{ +#ifdef EXT3_EXT_HAS_NO_TREE + struct ext3_ext_base *base = inode; +#else + struct ext3_extents_tree tree; + struct ext3_ext_base *base = &tree; + + ext3_init_tree_desc(base, inode); +#endif + /* for an ost_write request, it needs <#fragments> * + * metablocks at maxium b=16542 */ + *count = frags * (EXT_DEPTH(base) + 1) * EXT3_BLOCK_SIZE(sb); + return 0; +} + #endif lvfs_sbdev_type fsfilt_ext3_journal_sbdev(struct super_block *sb) @@ -2023,6 +2230,72 @@ lvfs_sbdev_type fsfilt_ext3_journal_sbdev(struct super_block *sb) } EXPORT_SYMBOL(fsfilt_ext3_journal_sbdev); +ssize_t lustre_read_quota(struct file *f, struct inode *inode, int type, + char *buf, int count, loff_t pos) +{ + loff_t p = pos; + int rc; + + if (!f && !inode) { + CERROR("lustre_read_quota failed for no quota file!\n"); + libcfs_debug_dumpstack(NULL); + return -EINVAL; + } + + /* Support for both adm and op quota files must be provided */ + if (f) { + rc = fsfilt_ext3_read_record(f, buf, count, &p); + rc = rc < 0 ? rc : p - pos; + } else { + struct super_block *sb = inode->i_sb; + rc = sb->s_op->quota_read(sb, type, buf, count, pos); + } + return rc; +} + +ssize_t lustre_write_quota(struct file *f, char *buf, int count, loff_t pos) +{ + loff_t p = pos; + int rc; + + /* Only adm quota files are supported, op updates are handled by vfs */ + rc = fsfilt_ext3_write_record(f, buf, count, &p, 0); + rc = rc < 0 ? rc : p - pos; + + return rc; +} + +void *lustre_quota_journal_start(struct inode *inode, int delete) +{ + handle_t *handle; + unsigned block_count; + + if (delete) { + /* each indirect block (+4) may become free, attaching to the + * header list of free blocks (+1); the data block (+1) may + * become a free block (+0) or a block with free dqentries (+0) */ + block_count = (4 + 1) + 1; + handle = ext3_journal_start(inode, + block_count*FSFILT_DATA_TRANS_BLOCKS(inode->i_sb)+2); + } else { + /* indirect blocks are touched (+4), each causes file expansion (+0) or + * freeblk reusage with a header update (+1); dqentry is either reused + * causing update of the entry block (+1), prev (+1) and next (+1) or + * a new block allocation (+1) with a header update (+1) */ + block_count = (4 + 1) + 3; + handle = ext3_journal_start(inode, + block_count*FSFILT_DATA_TRANS_BLOCKS(inode->i_sb)+2); + + } + + return handle; +} + +void lustre_quota_journal_stop(void *handle) +{ + ext3_journal_stop((handle_t *)handle); +} + static struct fsfilt_operations fsfilt_ext3_ops = { .fs_type = "ext3", .fs_owner = THIS_MODULE, @@ -2049,16 +2322,17 @@ static struct fsfilt_operations fsfilt_ext3_ops = { .fs_setup = fsfilt_ext3_setup, .fs_send_bio = fsfilt_ext3_send_bio, .fs_get_op_len = fsfilt_ext3_get_op_len, - .fs_quotactl = fsfilt_ext3_quotactl, - .fs_quotacheck = fsfilt_ext3_quotacheck, #ifdef HAVE_DISK_INODE_VERSION .fs_get_version = fsfilt_ext3_get_version, .fs_set_version = fsfilt_ext3_set_version, #endif #ifdef HAVE_QUOTA_SUPPORT + .fs_quotactl = fsfilt_ext3_quotactl, + .fs_quotacheck = fsfilt_ext3_quotacheck, .fs_quotainfo = fsfilt_ext3_quotainfo, .fs_qids = fsfilt_ext3_qids, .fs_dquot = fsfilt_ext3_dquot, + .fs_get_mblk = fsfilt_ext3_get_mblk, #endif .fs_journal_sbdev = fsfilt_ext3_journal_sbdev, }; @@ -2096,6 +2370,6 @@ static void __exit fsfilt_ext3_exit(void) module_init(fsfilt_ext3_init); module_exit(fsfilt_ext3_exit); -MODULE_AUTHOR("Cluster File Systems, Inc. "); +MODULE_AUTHOR("Sun Microsystems, Inc. "); MODULE_DESCRIPTION("Lustre ext3 Filesystem Helper v0.1"); MODULE_LICENSE("GPL");