X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Flvfs%2Ffsfilt_ext3.c;h=abdb74e2e5ee1812927b7c4f295d3fe7d28a87d8;hb=cf9e92a0adb490109f67d05da5833cfb270d92e8;hp=26e666e6eb5eb50fadd916975d943433c2ac9acb;hpb=9dce07db5dc5ff3e0ffa57289ac095966a88adcc;p=fs%2Flustre-release.git diff --git a/lustre/lvfs/fsfilt_ext3.c b/lustre/lvfs/fsfilt_ext3.c index 26e666e..abdb74e 100644 --- a/lustre/lvfs/fsfilt_ext3.c +++ b/lustre/lvfs/fsfilt_ext3.c @@ -26,7 +26,7 @@ * GPL HEADER END */ /* - * Copyright 2008 Sun Microsystems, Inc. All rights reserved + * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. */ /* @@ -57,9 +57,23 @@ #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 @@ -68,13 +82,11 @@ #include #include -#ifdef EXT3_MULTIBLOCK_ALLOCATOR #ifdef HAVE_EXT4_LDISKFS #include #else #include #endif -#endif #include "lustre_quota_fmt.h" @@ -93,8 +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; @@ -123,6 +141,18 @@ struct fsfilt_cb_data { #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; @@ -163,14 +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) { CDEBUG(D_INFO, "Get version "LPX64" for inode %lu\n", - EXT3_I(inode)->i_fs_version, inode->i_ino); - return EXT3_I(inode)->i_fs_version; + get_i_version(inode), inode->i_ino); + return get_i_version(inode); } /* @@ -178,11 +227,11 @@ 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; + __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); - (EXT3_I(inode))->i_fs_version = new_version; + 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); @@ -273,19 +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(); } @@ -300,7 +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); @@ -386,7 +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 @@ -437,7 +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)); @@ -477,7 +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; } @@ -498,7 +534,7 @@ 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; @@ -555,7 +591,7 @@ 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) && + if (!cfs_curproc_is_in_groups(inode->i_gid) && !cfs_capable(CFS_CAP_FSETID)) inode->i_mode &= ~S_ISGID; } @@ -593,9 +629,21 @@ static int fsfilt_ext3_iocontrol(struct inode *inode, struct file *file, 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 @@ -673,8 +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; @@ -776,26 +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 fsfilt_up_truncate_sem(inode) up(&EXT3_I(inode)->truncate_sem); -#define fsfilt_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 -#ifdef HAVE_EXT4_LDISKFS -#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)); -#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 +# 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 @@ -832,8 +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; @@ -973,7 +1021,7 @@ static int ext3_ext_new_extent_cb(struct ext3_ext_base *base, count = ext3_ext_calc_credits_for_insert(base, path); 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)) { fsfilt_down_truncate_sem(inode); return PTR_ERR(handle); @@ -982,7 +1030,7 @@ static int ext3_ext_new_extent_cb(struct ext3_ext_base *base, 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; } @@ -996,7 +1044,7 @@ static int ext3_ext_new_extent_cb(struct ext3_ext_base *base, nex.ee_block = cpu_to_le32(cex->ec_block); ext3_ext_store_pblock(&nex, pblock); nex.ee_len = cpu_to_le16(count); - err = ext3_ext_insert_extent(handle, base, path, &nex); + err = fsfilt_ext3_ext_insert_extent(handle, base, path, &nex, 0); if (err) { /* free data blocks we just allocated */ /* not a good idea to call discard here directly, @@ -1020,7 +1068,7 @@ static int ext3_ext_new_extent_cb(struct ext3_ext_base *base, 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 */ @@ -1143,7 +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); @@ -1172,22 +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; } @@ -1199,19 +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; @@ -1294,14 +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) @@ -1322,7 +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", @@ -1335,7 +1381,7 @@ 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; } @@ -1465,15 +1511,10 @@ static int fsfilt_ext3_quotactl(struct super_block *sb, LASSERT(oqc->qc_id == LUSTRE_QUOTA_V2); - if (!qcop->quota_on) - GOTO(out, rc = -ENOSYS); - - rc = qcop->quota_on(sb, i, QFMT_VFS_V0, - name[i]); + 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) @@ -1551,8 +1592,8 @@ out: } struct chk_dqblk{ - struct hlist_node dqb_hash; /** quotacheck hash */ - struct list_head dqb_list; /** in list also */ + 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 */ @@ -1575,13 +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; } @@ -1595,8 +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; } @@ -1605,10 +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; @@ -1633,8 +1674,8 @@ 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; } @@ -1678,8 +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; @@ -1692,44 +1734,30 @@ get_group_desc(struct super_block *sb, int group) return gdp + desc; } - -#ifndef HAVE_EXT4_LDISKFS static inline struct buffer_head * 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); + desc = get_group_desc(sb, group, NULL); bh = sb_bread(sb, ext3_inode_bitmap(sb, desc)); return bh; } -#endif -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)) -#ifdef HAVE_EXT4_LDISKFS - inode = ext4_iget(sb, ino); - if (IS_ERR(inode)) - /* Newer kernels return an error instead of a NULL pointer */ - inode = NULL; +static __u32 ext3_itable_unused_count(struct super_block *sb, + struct ext3_group_desc *bg) { + return le16_to_cpu(bg->bg_itable_unused); +} #else - inode = iget(sb, ino); +#define get_group_desc ext3_get_group_desc #endif - return inode; -} 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, @@ -1915,28 +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; @@ -1949,8 +1981,8 @@ 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)) @@ -1967,21 +1999,63 @@ static int fsfilt_ext3_quotacheck(struct super_block *sb, 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 = ext3_read_inode_bitmap(sb, group); if (!bitmap_bh) { - CERROR("ext3_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) { @@ -1989,7 +2063,6 @@ static int fsfilt_ext3_quotacheck(struct super_block *sb, GOTO(out, rc); } } - brelse(bitmap_bh); } @@ -1997,7 +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)) @@ -2008,7 +2081,7 @@ static int fsfilt_ext3_quotacheck(struct super_block *sb, LASSERT(sb_dqopt(sb)->files[i] != NULL); - INIT_LIST_HEAD(&id_list); + 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 @@ -2017,8 +2090,8 @@ 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, @@ -2061,7 +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) { @@ -2092,7 +2165,7 @@ static int fsfilt_ext3_quotainfo(struct lustre_quota_info *lqi, int type, } 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); } @@ -2104,7 +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) { @@ -2116,9 +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) @@ -2157,6 +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,