X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=fs%2Fnfs%2Fdir.c;h=7422b734fd096782aace396228ed2041b9d29aec;hb=97bf2856c6014879bd04983a3e9dfcdac1e7fe85;hp=f061e70cdc7dd59a18dc9e054321a57ab1d66cfd;hpb=9bf4aaab3e101692164d49b7ca357651eb691cb6;p=linux-2.6.git diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index f061e70cd..7422b734f 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -30,7 +30,14 @@ #include #include #include +#include #include +#include +#include + +#include "nfs4_fs.h" +#include "delegation.h" +#include "iostat.h" #define NFS_PARANOIA 1 /* #define NFS_DEBUG_VERBOSE 1 */ @@ -38,8 +45,6 @@ static int nfs_opendir(struct inode *, struct file *); static int nfs_readdir(struct file *, void *, filldir_t); static struct dentry *nfs_lookup(struct inode *, struct dentry *, struct nameidata *); -static int nfs_cached_lookup(struct inode *, struct dentry *, - struct nfs_fh *, struct nfs_fattr *); static int nfs_create(struct inode *, struct dentry *, int, struct nameidata *); static int nfs_mkdir(struct inode *, struct dentry *, int); static int nfs_rmdir(struct inode *, struct dentry *); @@ -49,12 +54,16 @@ static int nfs_link(struct dentry *, struct inode *, struct dentry *); static int nfs_mknod(struct inode *, struct dentry *, int, dev_t); static int nfs_rename(struct inode *, struct dentry *, struct inode *, struct dentry *); +static int nfs_fsync_dir(struct file *, struct dentry *, int); +static loff_t nfs_llseek_dir(struct file *, loff_t, int); -struct file_operations nfs_dir_operations = { +const struct file_operations nfs_dir_operations = { + .llseek = nfs_llseek_dir, .read = generic_read_dir, .readdir = nfs_readdir, .open = nfs_opendir, .release = nfs_release, + .fsync = nfs_fsync_dir, }; struct inode_operations nfs_dir_inode_operations = { @@ -72,6 +81,27 @@ struct inode_operations nfs_dir_inode_operations = { .setattr = nfs_setattr, }; +#ifdef CONFIG_NFS_V3 +struct inode_operations nfs3_dir_inode_operations = { + .create = nfs_create, + .lookup = nfs_lookup, + .link = nfs_link, + .unlink = nfs_unlink, + .symlink = nfs_symlink, + .mkdir = nfs_mkdir, + .rmdir = nfs_rmdir, + .mknod = nfs_mknod, + .rename = nfs_rename, + .permission = nfs_permission, + .getattr = nfs_getattr, + .setattr = nfs_setattr, + .listxattr = nfs3_listxattr, + .getxattr = nfs3_getxattr, + .setxattr = nfs3_setxattr, + .removexattr = nfs3_removexattr, +}; +#endif /* CONFIG_NFS_V3 */ + #ifdef CONFIG_NFS_V4 static struct dentry *nfs_atomic_lookup(struct inode *, struct dentry *, struct nameidata *); @@ -88,6 +118,9 @@ struct inode_operations nfs4_dir_inode_operations = { .permission = nfs_permission, .getattr = nfs_getattr, .setattr = nfs_setattr, + .getxattr = nfs4_getxattr, + .setxattr = nfs4_setxattr, + .listxattr = nfs4_listxattr, }; #endif /* CONFIG_NFS_V4 */ @@ -98,23 +131,26 @@ struct inode_operations nfs4_dir_inode_operations = { static int nfs_opendir(struct inode *inode, struct file *filp) { - int res = 0; + int res; + + dfprintk(VFS, "NFS: opendir(%s/%ld)\n", + inode->i_sb->s_id, inode->i_ino); lock_kernel(); /* Call generic open code in order to cache credentials */ - if (!res) - res = nfs_open(inode, filp); + res = nfs_open(inode, filp); unlock_kernel(); return res; } -typedef u32 * (*decode_dirent_t)(u32 *, struct nfs_entry *, int); +typedef __be32 * (*decode_dirent_t)(__be32 *, struct nfs_entry *, int); typedef struct { struct file *file; struct page *page; unsigned long page_index; - u32 *ptr; - u64 target; + __be32 *ptr; + u64 *dir_cookie; + loff_t current_index; struct nfs_entry *entry; decode_dirent_t decode; int plus; @@ -137,36 +173,40 @@ static int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page *page) { struct file *file = desc->file; - struct inode *inode = file->f_dentry->d_inode; + struct inode *inode = file->f_path.dentry->d_inode; struct rpc_cred *cred = nfs_file_cred(file); unsigned long timestamp; int error; - dfprintk(VFS, "NFS: nfs_readdir_filler() reading cookie %Lu into page %lu.\n", (long long)desc->entry->cookie, page->index); + dfprintk(DIRCACHE, "NFS: %s: reading cookie %Lu into page %lu\n", + __FUNCTION__, (long long)desc->entry->cookie, + page->index); again: timestamp = jiffies; - error = NFS_PROTO(inode)->readdir(file->f_dentry, cred, desc->entry->cookie, page, + error = NFS_PROTO(inode)->readdir(file->f_path.dentry, cred, desc->entry->cookie, page, NFS_SERVER(inode)->dtsize, desc->plus); if (error < 0) { /* We requested READDIRPLUS, but the server doesn't grok it */ if (error == -ENOTSUPP && desc->plus) { NFS_SERVER(inode)->caps &= ~NFS_CAP_READDIRPLUS; - NFS_FLAGS(inode) &= ~NFS_INO_ADVISE_RDPLUS; + clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_FLAGS(inode)); desc->plus = 0; goto again; } goto error; } SetPageUptodate(page); - NFS_FLAGS(inode) |= NFS_INO_INVALID_ATIME; + spin_lock(&inode->i_lock); + NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATIME; + spin_unlock(&inode->i_lock); /* Ensure consistent page alignment of the data. * Note: assumes we have exclusive access to this mapping either - * throught inode->i_sem or some other mechanism. + * through inode->i_mutex or some other mechanism. */ - if (page->index == 0) { - invalidate_inode_pages(inode->i_mapping); - NFS_I(inode)->readdir_timestamp = timestamp; + if (page->index == 0 && invalidate_inode_pages2_range(inode->i_mapping, PAGE_CACHE_SIZE, -1) < 0) { + /* Should never happen */ + nfs_zap_mapping(inode, inode->i_mapping); } unlock_page(page); return 0; @@ -181,7 +221,7 @@ int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page *page) static inline int dir_decode(nfs_readdir_descriptor_t *desc) { - u32 *p = desc->ptr; + __be32 *p = desc->ptr; p = desc->decode(p, desc->entry, desc->plus); if (IS_ERR(p)) return PTR_ERR(p); @@ -200,44 +240,82 @@ void dir_page_release(nfs_readdir_descriptor_t *desc) /* * Given a pointer to a buffer that has already been filled by a call - * to readdir, find the next entry. + * to readdir, find the next entry with cookie '*desc->dir_cookie'. * * If the end of the buffer has been reached, return -EAGAIN, if not, * return the offset within the buffer of the next entry to be * read. */ static inline -int find_dirent(nfs_readdir_descriptor_t *desc, struct page *page) +int find_dirent(nfs_readdir_descriptor_t *desc) { struct nfs_entry *entry = desc->entry; int loop_count = 0, status; while((status = dir_decode(desc)) == 0) { - dfprintk(VFS, "NFS: found cookie %Lu\n", (long long)entry->cookie); - if (entry->prev_cookie == desc->target) + dfprintk(DIRCACHE, "NFS: %s: examining cookie %Lu\n", + __FUNCTION__, (unsigned long long)entry->cookie); + if (entry->prev_cookie == *desc->dir_cookie) + break; + if (loop_count++ > 200) { + loop_count = 0; + schedule(); + } + } + return status; +} + +/* + * Given a pointer to a buffer that has already been filled by a call + * to readdir, find the entry at offset 'desc->file->f_pos'. + * + * If the end of the buffer has been reached, return -EAGAIN, if not, + * return the offset within the buffer of the next entry to be + * read. + */ +static inline +int find_dirent_index(nfs_readdir_descriptor_t *desc) +{ + struct nfs_entry *entry = desc->entry; + int loop_count = 0, + status; + + for(;;) { + status = dir_decode(desc); + if (status) + break; + + dfprintk(DIRCACHE, "NFS: found cookie %Lu at index %Ld\n", + (unsigned long long)entry->cookie, desc->current_index); + + if (desc->file->f_pos == desc->current_index) { + *desc->dir_cookie = entry->cookie; break; + } + desc->current_index++; if (loop_count++ > 200) { loop_count = 0; schedule(); } } - dfprintk(VFS, "NFS: find_dirent() returns %d\n", status); return status; } /* - * Find the given page, and call find_dirent() in order to try to - * return the next entry. + * Find the given page, and call find_dirent() or find_dirent_index in + * order to try to return the next entry. */ static inline int find_dirent_page(nfs_readdir_descriptor_t *desc) { - struct inode *inode = desc->file->f_dentry->d_inode; + struct inode *inode = desc->file->f_path.dentry->d_inode; struct page *page; int status; - dfprintk(VFS, "NFS: find_dirent_page() searching directory page %ld\n", desc->page_index); + dfprintk(DIRCACHE, "NFS: %s: searching page %ld for target %Lu\n", + __FUNCTION__, desc->page_index, + (long long) *desc->dir_cookie); page = read_cache_page(inode->i_mapping, desc->page_index, (filler_t *)nfs_readdir_filler, desc); @@ -251,11 +329,14 @@ int find_dirent_page(nfs_readdir_descriptor_t *desc) /* NOTE: Someone else may have changed the READDIRPLUS flag */ desc->page = page; desc->ptr = kmap(page); /* matching kunmap in nfs_do_filldir */ - status = find_dirent(desc, page); + if (*desc->dir_cookie != 0) + status = find_dirent(desc); + else + status = find_dirent_index(desc); if (status < 0) dir_page_release(desc); out: - dfprintk(VFS, "NFS: find_dirent_page() returns %d\n", status); + dfprintk(DIRCACHE, "NFS: %s: returns %d\n", __FUNCTION__, status); return status; read_error: page_cache_release(page); @@ -266,7 +347,8 @@ int find_dirent_page(nfs_readdir_descriptor_t *desc) * Recurse through the page cache pages, and return a * filled nfs_entry structure of the next directory entry if possible. * - * The target for the search is 'desc->target'. + * The target for the search is '*desc->dir_cookie' if non-0, + * 'desc->file->f_pos' otherwise */ static inline int readdir_search_pagecache(nfs_readdir_descriptor_t *desc) @@ -274,7 +356,18 @@ int readdir_search_pagecache(nfs_readdir_descriptor_t *desc) int loop_count = 0; int res; - dfprintk(VFS, "NFS: readdir_search_pagecache() searching for cookie %Lu\n", (long long)desc->target); + /* Always search-by-index from the beginning of the cache */ + if (*desc->dir_cookie == 0) { + dfprintk(DIRCACHE, "NFS: readdir_search_pagecache() searching for offset %Ld\n", + (long long)desc->file->f_pos); + desc->page_index = 0; + desc->entry->cookie = desc->entry->prev_cookie = 0; + desc->entry->eof = 0; + desc->current_index = 0; + } else + dfprintk(DIRCACHE, "NFS: readdir_search_pagecache() searching for cookie %Lu\n", + (unsigned long long)*desc->dir_cookie); + for (;;) { res = find_dirent_page(desc); if (res != -EAGAIN) @@ -286,28 +379,18 @@ int readdir_search_pagecache(nfs_readdir_descriptor_t *desc) schedule(); } } - dfprintk(VFS, "NFS: readdir_search_pagecache() returned %d\n", res); + + dfprintk(DIRCACHE, "NFS: %s: returns %d\n", __FUNCTION__, res); return res; } -static unsigned int nfs_type2dtype[] = { - DT_UNKNOWN, - DT_REG, - DT_DIR, - DT_BLK, - DT_CHR, - DT_LNK, - DT_SOCK, - DT_UNKNOWN, - DT_FIFO -}; - -static inline -unsigned int nfs_type_to_d_type(enum nfs_ftype type) +static inline unsigned int dt_type(struct inode *inode) { - return nfs_type2dtype[type]; + return (inode->i_mode >> 12) & 15; } +static struct dentry *nfs_readdir_lookup(nfs_readdir_descriptor_t *desc); + /* * Once we've found the start of the dirent within a page: fill 'er up... */ @@ -317,11 +400,13 @@ int nfs_do_filldir(nfs_readdir_descriptor_t *desc, void *dirent, { struct file *file = desc->file; struct nfs_entry *entry = desc->entry; + struct dentry *dentry = NULL; unsigned long fileid; int loop_count = 0, res; - dfprintk(VFS, "NFS: nfs_do_filldir() filling starting @ cookie %Lu\n", (long long)desc->target); + dfprintk(DIRCACHE, "NFS: nfs_do_filldir() filling starting @ cookie %Lu\n", + (unsigned long long)entry->cookie); for(;;) { unsigned d_type = DT_UNKNOWN; @@ -329,15 +414,23 @@ int nfs_do_filldir(nfs_readdir_descriptor_t *desc, void *dirent, * retrieving the current dirent on the server */ fileid = nfs_fileid_to_ino_t(entry->ino); + /* Get a dentry if we have one */ + if (dentry != NULL) + dput(dentry); + dentry = nfs_readdir_lookup(desc); + /* Use readdirplus info */ - if (desc->plus && (entry->fattr->valid & NFS_ATTR_FATTR)) - d_type = nfs_type_to_d_type(entry->fattr->type); + if (dentry != NULL && dentry->d_inode != NULL) { + d_type = dt_type(dentry->d_inode); + fileid = dentry->d_inode->i_ino; + } res = filldir(dirent, entry->name, entry->len, - entry->prev_cookie, fileid, d_type); + file->f_pos, fileid, d_type); if (res < 0) break; - file->f_pos = desc->target = entry->cookie; + file->f_pos++; + *desc->dir_cookie = entry->cookie; if (dir_decode(desc) != 0) { desc->page_index ++; break; @@ -348,8 +441,10 @@ int nfs_do_filldir(nfs_readdir_descriptor_t *desc, void *dirent, } } dir_page_release(desc); - - dfprintk(VFS, "NFS: nfs_do_filldir() filling ended @ cookie %Lu; returning = %d\n", (long long)desc->target, res); + if (dentry != NULL) + dput(dentry); + dfprintk(DIRCACHE, "NFS: nfs_do_filldir() filling ended @ cookie %Lu; returning = %d\n", + (unsigned long long)*desc->dir_cookie, res); return res; } @@ -370,28 +465,31 @@ int uncached_readdir(nfs_readdir_descriptor_t *desc, void *dirent, filldir_t filldir) { struct file *file = desc->file; - struct inode *inode = file->f_dentry->d_inode; + struct inode *inode = file->f_path.dentry->d_inode; struct rpc_cred *cred = nfs_file_cred(file); struct page *page = NULL; int status; - dfprintk(VFS, "NFS: uncached_readdir() searching for cookie %Lu\n", (long long)desc->target); + dfprintk(DIRCACHE, "NFS: uncached_readdir() searching for cookie %Lu\n", + (unsigned long long)*desc->dir_cookie); page = alloc_page(GFP_HIGHUSER); if (!page) { status = -ENOMEM; goto out; } - desc->error = NFS_PROTO(inode)->readdir(file->f_dentry, cred, desc->target, + desc->error = NFS_PROTO(inode)->readdir(file->f_path.dentry, cred, *desc->dir_cookie, page, NFS_SERVER(inode)->dtsize, desc->plus); - NFS_FLAGS(inode) |= NFS_INO_INVALID_ATIME; + spin_lock(&inode->i_lock); + NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATIME; + spin_unlock(&inode->i_lock); desc->page = page; desc->ptr = kmap(page); /* matching kunmap in nfs_do_filldir */ if (desc->error >= 0) { if ((status = dir_decode(desc)) == 0) - desc->entry->prev_cookie = desc->target; + desc->entry->prev_cookie = *desc->dir_cookie; } else status = -EIO; if (status < 0) @@ -405,19 +503,21 @@ int uncached_readdir(nfs_readdir_descriptor_t *desc, void *dirent, desc->entry->cookie = desc->entry->prev_cookie = 0; desc->entry->eof = 0; out: - dfprintk(VFS, "NFS: uncached_readdir() returns %d\n", status); + dfprintk(DIRCACHE, "NFS: %s: returns %d\n", + __FUNCTION__, status); return status; out_release: dir_page_release(desc); goto out; } -/* The file offset position is now represented as a true offset into the - * page cache as is the case in most of the other filesystems. +/* The file offset position represents the dirent entry number. A + last cookie cache takes care of the common case of reading the + whole directory. */ static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir) { - struct dentry *dentry = filp->f_dentry; + struct dentry *dentry = filp->f_path.dentry; struct inode *inode = dentry->d_inode; nfs_readdir_descriptor_t my_desc, *desc = &my_desc; @@ -426,24 +526,29 @@ static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir) struct nfs_fattr fattr; long res; + dfprintk(VFS, "NFS: readdir(%s/%s) starting at cookie %Lu\n", + dentry->d_parent->d_name.name, dentry->d_name.name, + (long long)filp->f_pos); + nfs_inc_stats(inode, NFSIOS_VFSGETDENTS); + lock_kernel(); - res = nfs_revalidate_inode(NFS_SERVER(inode), inode); + res = nfs_revalidate_mapping_nolock(inode, filp->f_mapping); if (res < 0) { unlock_kernel(); return res; } /* - * filp->f_pos points to the file offset in the page cache. - * but if the cache has meanwhile been zapped, we need to - * read from the last dirent to revalidate f_pos - * itself. + * filp->f_pos points to the dirent entry number. + * *desc->dir_cookie has the cookie for the next entry. We have + * to either find the entry with the appropriate number or + * revalidate the cookie. */ memset(desc, 0, sizeof(*desc)); desc->file = filp; - desc->target = filp->f_pos; + desc->dir_cookie = &((struct nfs_open_context *)filp->private_data)->dir_cookie; desc->decode = NFS_PROTO(inode)->decode_dirent; desc->plus = NFS_USE_READDIRPLUS(inode); @@ -451,13 +556,15 @@ static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir) my_entry.eof = 0; my_entry.fh = &fh; my_entry.fattr = &fattr; + nfs_fattr_init(&fattr); desc->entry = &my_entry; while(!desc->entry->eof) { res = readdir_search_pagecache(desc); + if (res == -EBADCOOKIE) { /* This means either end of directory */ - if (desc->entry->cookie != desc->target) { + if (*desc->dir_cookie && desc->entry->cookie != *desc->dir_cookie) { /* Or that the server has 'lost' a cookie */ res = uncached_readdir(desc, dirent, filldir); if (res >= 0) @@ -467,7 +574,7 @@ static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir) break; } if (res == -ETOOSMALL && desc->plus) { - NFS_FLAGS(inode) &= ~NFS_INO_ADVISE_RDPLUS; + clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_FLAGS(inode)); nfs_zap_caches(inode); desc->plus = 0; desc->entry->eof = 0; @@ -483,10 +590,46 @@ static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir) } } unlock_kernel(); - if (desc->error < 0) - return desc->error; - if (res < 0) - return res; + if (res > 0) + res = 0; + dfprintk(VFS, "NFS: readdir(%s/%s) returns %ld\n", + dentry->d_parent->d_name.name, dentry->d_name.name, + res); + return res; +} + +loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int origin) +{ + mutex_lock(&filp->f_path.dentry->d_inode->i_mutex); + switch (origin) { + case 1: + offset += filp->f_pos; + case 0: + if (offset >= 0) + break; + default: + offset = -EINVAL; + goto out; + } + if (offset != filp->f_pos) { + filp->f_pos = offset; + ((struct nfs_open_context *)filp->private_data)->dir_cookie = 0; + } +out: + mutex_unlock(&filp->f_path.dentry->d_inode->i_mutex); + return offset; +} + +/* + * All directory operations under NFS are synchronous, so fsync() + * is a dummy operation. + */ +int nfs_fsync_dir(struct file *filp, struct dentry *dentry, int datasync) +{ + dfprintk(VFS, "NFS: fsync_dir(%s/%s) datasync %d\n", + dentry->d_parent->d_name.name, dentry->d_name.name, + datasync); + return 0; } @@ -499,7 +642,7 @@ static inline int nfs_check_verifier(struct inode *dir, struct dentry *dentry) { if (IS_ROOT(dentry)) return 1; - if ((NFS_FLAGS(dir) & NFS_INO_INVALID_ATTR) != 0 + if ((NFS_I(dir)->cache_validity & NFS_INO_INVALID_ATTR) != 0 || nfs_attribute_timeout(dir)) return 0; return nfs_verify_change_attribute(dir, (unsigned long)dentry->d_fsdata); @@ -519,14 +662,47 @@ static inline void nfs_renew_times(struct dentry * dentry) dentry->d_time = jiffies; } +/* + * Return the intent data that applies to this particular path component + * + * Note that the current set of intents only apply to the very last + * component of the path. + * We check for this using LOOKUP_CONTINUE and LOOKUP_PARENT. + */ +static inline unsigned int nfs_lookup_check_intent(struct nameidata *nd, unsigned int mask) +{ + if (nd->flags & (LOOKUP_CONTINUE|LOOKUP_PARENT)) + return 0; + return nd->flags & mask; +} + +/* + * Inode and filehandle revalidation for lookups. + * + * We force revalidation in the cases where the VFS sets LOOKUP_REVAL, + * or if the intent information indicates that we're about to open this + * particular file and the "nocto" mount flag is not set. + * + */ static inline -int nfs_lookup_verify_inode(struct inode *inode, int isopen) +int nfs_lookup_verify_inode(struct inode *inode, struct nameidata *nd) { struct nfs_server *server = NFS_SERVER(inode); - if (isopen && !(server->flags & NFS_MOUNT_NOCTO)) - return __nfs_revalidate_inode(server, inode); + if (nd != NULL) { + /* VFS wants an on-the-wire revalidation */ + if (nd->flags & LOOKUP_REVAL) + goto out_force; + /* This is an open(2) */ + if (nfs_lookup_check_intent(nd, LOOKUP_OPEN) != 0 && + !(server->flags & NFS_MOUNT_NOCTO) && + (S_ISREG(inode->i_mode) || + S_ISDIR(inode->i_mode))) + goto out_force; + } return nfs_revalidate_inode(server, inode); +out_force: + return __nfs_revalidate_inode(server, inode); } /* @@ -540,12 +716,8 @@ static inline int nfs_neg_need_reval(struct inode *dir, struct dentry *dentry, struct nameidata *nd) { - int ndflags = 0; - - if (nd) - ndflags = nd->flags; /* Don't revalidate a negative dentry if we're creating a new file */ - if ((ndflags & LOOKUP_CREATE) && !(ndflags & LOOKUP_CONTINUE)) + if (nd != NULL && nfs_lookup_check_intent(nd, LOOKUP_CREATE) != 0) return 0; return !nfs_check_verifier(dir, dentry); } @@ -570,16 +742,13 @@ static int nfs_lookup_revalidate(struct dentry * dentry, struct nameidata *nd) struct nfs_fh fhandle; struct nfs_fattr fattr; unsigned long verifier; - int isopen = 0; parent = dget_parent(dentry); lock_kernel(); dir = parent->d_inode; + nfs_inc_stats(dir, NFSIOS_DENTRYREVALIDATE); inode = dentry->d_inode; - if (nd && !(nd->flags & LOOKUP_CONTINUE) && (nd->flags & LOOKUP_OPEN)) - isopen = 1; - if (!inode) { if (nfs_neg_need_reval(dir, dentry, nd)) goto out_bad; @@ -587,53 +756,43 @@ static int nfs_lookup_revalidate(struct dentry * dentry, struct nameidata *nd) } if (is_bad_inode(inode)) { - dfprintk(VFS, "nfs_lookup_validate: %s/%s has dud inode\n", - dentry->d_parent->d_name.name, dentry->d_name.name); + dfprintk(LOOKUPCACHE, "%s: %s/%s has dud inode\n", + __FUNCTION__, dentry->d_parent->d_name.name, + dentry->d_name.name); goto out_bad; } /* Revalidate parent directory attribute cache */ - nfs_revalidate_inode(NFS_SERVER(dir), dir); + if (nfs_revalidate_inode(NFS_SERVER(dir), dir) < 0) + goto out_zap_parent; /* Force a full look up iff the parent directory has changed */ if (nfs_check_verifier(dir, dentry)) { - if (nfs_lookup_verify_inode(inode, isopen)) + if (nfs_lookup_verify_inode(inode, nd)) goto out_zap_parent; goto out_valid; } - /* - * Note: we're not holding inode->i_sem and so may be racing with - * operations that change the directory. We therefore save the - * change attribute *before* we do the RPC call. - */ - verifier = nfs_save_change_attribute(dir); - error = nfs_cached_lookup(dir, dentry, &fhandle, &fattr); - if (!error) { - if (memcmp(NFS_FH(inode), &fhandle, sizeof(struct nfs_fh))!= 0) - goto out_bad; - if (nfs_lookup_verify_inode(inode, isopen)) - goto out_zap_parent; - goto out_valid_renew; - } - if (NFS_STALE(inode)) goto out_bad; + verifier = nfs_save_change_attribute(dir); error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, &fhandle, &fattr); if (error) goto out_bad; - if (memcmp(NFS_FH(inode), &fhandle, sizeof(struct nfs_fh))!= 0) + if (nfs_compare_fh(NFS_FH(inode), &fhandle)) goto out_bad; if ((error = nfs_refresh_inode(inode, &fattr)) != 0) goto out_bad; - out_valid_renew: nfs_renew_times(dentry); nfs_set_verifier(dentry, verifier); out_valid: unlock_kernel(); dput(parent); + dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is valid\n", + __FUNCTION__, dentry->d_parent->d_name.name, + dentry->d_name.name); return 1; out_zap_parent: nfs_zap_caches(dir); @@ -650,6 +809,9 @@ out_zap_parent: d_drop(dentry); unlock_kernel(); dput(parent); + dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is invalid\n", + __FUNCTION__, dentry->d_parent->d_name.name, + dentry->d_name.name); return 0; } @@ -681,9 +843,10 @@ static int nfs_dentry_delete(struct dentry *dentry) */ static void nfs_dentry_iput(struct dentry *dentry, struct inode *inode) { + nfs_inode_return_delegation(inode); if (dentry->d_flags & DCACHE_NFSFS_RENAMED) { lock_kernel(); - inode->i_nlink--; + drop_nlink(inode); nfs_complete_unlink(dentry); unlock_kernel(); } @@ -698,18 +861,34 @@ struct dentry_operations nfs_dentry_operations = { .d_iput = nfs_dentry_iput, }; +/* + * Use intent information to check whether or not we're going to do + * an O_EXCL create using this path component. + */ static inline int nfs_is_exclusive_create(struct inode *dir, struct nameidata *nd) { if (NFS_PROTO(dir)->version == 2) return 0; - if (!nd || (nd->flags & LOOKUP_CONTINUE) || !(nd->flags & LOOKUP_CREATE)) + if (nd == NULL || nfs_lookup_check_intent(nd, LOOKUP_CREATE) == 0) return 0; return (nd->intent.open.flags & O_EXCL) != 0; } +static inline int nfs_reval_fsid(struct vfsmount *mnt, struct inode *dir, + struct nfs_fh *fh, struct nfs_fattr *fattr) +{ + struct nfs_server *server = NFS_SERVER(dir); + + if (!nfs_fsid_equal(&server->fsid, &fattr->fsid)) + /* Revalidate fsid on root dir */ + return __nfs_revalidate_inode(server, mnt->mnt_root->d_inode); + return 0; +} + static struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd) { + struct dentry *res; struct inode *inode = NULL; int error; struct nfs_fh fhandle; @@ -717,45 +896,64 @@ static struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, stru dfprintk(VFS, "NFS: lookup(%s/%s)\n", dentry->d_parent->d_name.name, dentry->d_name.name); + nfs_inc_stats(dir, NFSIOS_VFSLOOKUP); - error = -ENAMETOOLONG; + res = ERR_PTR(-ENAMETOOLONG); if (dentry->d_name.len > NFS_SERVER(dir)->namelen) goto out; - error = -ENOMEM; + res = ERR_PTR(-ENOMEM); dentry->d_op = NFS_PROTO(dir)->dentry_ops; lock_kernel(); - /* Revalidate parent directory attribute cache */ - nfs_revalidate_inode(NFS_SERVER(dir), dir); - /* If we're doing an exclusive create, optimize away the lookup */ - if (nfs_is_exclusive_create(dir, nd)) - goto no_entry; + /* + * If we're doing an exclusive create, optimize away the lookup + * but don't hash the dentry. + */ + if (nfs_is_exclusive_create(dir, nd)) { + d_instantiate(dentry, NULL); + res = NULL; + goto out_unlock; + } - error = nfs_cached_lookup(dir, dentry, &fhandle, &fattr); - if (error != 0) { - error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, - &fhandle, &fattr); - if (error == -ENOENT) - goto no_entry; - if (error != 0) - goto out_unlock; + error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, &fhandle, &fattr); + if (error == -ENOENT) + goto no_entry; + if (error < 0) { + res = ERR_PTR(error); + goto out_unlock; + } + error = nfs_reval_fsid(nd->mnt, dir, &fhandle, &fattr); + if (error < 0) { + res = ERR_PTR(error); + goto out_unlock; } - error = -EACCES; inode = nfs_fhget(dentry->d_sb, &fhandle, &fattr); - if (!inode) + res = (struct dentry *)inode; + if (IS_ERR(res)) goto out_unlock; + + dx_propagate_tag(nd, inode); no_entry: - error = 0; - d_add(dentry, inode); + res = d_materialise_unique(dentry, inode); + if (res != NULL) { + struct dentry *parent; + if (IS_ERR(res)) + goto out_unlock; + /* Was a directory renamed! */ + parent = dget_parent(res); + if (!IS_ROOT(parent)) + nfs_mark_for_revalidate(parent->d_inode); + dput(parent); + dentry = res; + } nfs_renew_times(dentry); nfs_set_verifier(dentry, nfs_save_change_attribute(dir)); out_unlock: unlock_kernel(); out: - BUG_ON(error > 0); - return ERR_PTR(error); + return res; } #ifdef CONFIG_NFS_V4 @@ -767,77 +965,89 @@ struct dentry_operations nfs4_dentry_operations = { .d_iput = nfs_dentry_iput, }; +/* + * Use intent information to determine whether we need to substitute + * the NFSv4-style stateful OPEN for the LOOKUP call + */ static int is_atomic_open(struct inode *dir, struct nameidata *nd) { - if (!nd) - return 0; - /* Check that we are indeed trying to open this file */ - if ((nd->flags & LOOKUP_CONTINUE) || !(nd->flags & LOOKUP_OPEN)) + if (nd == NULL || nfs_lookup_check_intent(nd, LOOKUP_OPEN) == 0) return 0; /* NFS does not (yet) have a stateful open for directories */ if (nd->flags & LOOKUP_DIRECTORY) return 0; /* Are we trying to write to a read only partition? */ - if (IS_RDONLY(dir) && (nd->intent.open.flags & (O_CREAT|O_TRUNC|FMODE_WRITE))) + if ((IS_RDONLY(dir) || MNT_IS_RDONLY(nd->mnt)) && + (nd->intent.open.flags & (O_CREAT|O_TRUNC|FMODE_WRITE))) return 0; return 1; } static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd) { - struct inode *inode = NULL; - int error = 0; + struct dentry *res = NULL; + int error; + + dfprintk(VFS, "NFS: atomic_lookup(%s/%ld), %s\n", + dir->i_sb->s_id, dir->i_ino, dentry->d_name.name); /* Check that we are indeed trying to open this file */ if (!is_atomic_open(dir, nd)) goto no_open; if (dentry->d_name.len > NFS_SERVER(dir)->namelen) { - error = -ENAMETOOLONG; + res = ERR_PTR(-ENAMETOOLONG); goto out; } dentry->d_op = NFS_PROTO(dir)->dentry_ops; /* Let vfs_create() deal with O_EXCL */ - if (nd->intent.open.flags & O_EXCL) - goto no_entry; + if (nd->intent.open.flags & O_EXCL) { + d_add(dentry, NULL); + goto out; + } /* Open the file on the server */ lock_kernel(); /* Revalidate parent directory attribute cache */ - nfs_revalidate_inode(NFS_SERVER(dir), dir); + error = nfs_revalidate_inode(NFS_SERVER(dir), dir); + if (error < 0) { + res = ERR_PTR(error); + unlock_kernel(); + goto out; + } if (nd->intent.open.flags & O_CREAT) { nfs_begin_data_update(dir); - inode = nfs4_atomic_open(dir, dentry, nd); + res = nfs4_atomic_open(dir, dentry, nd); nfs_end_data_update(dir); } else - inode = nfs4_atomic_open(dir, dentry, nd); + res = nfs4_atomic_open(dir, dentry, nd); unlock_kernel(); - if (IS_ERR(inode)) { - error = PTR_ERR(inode); + if (IS_ERR(res)) { + error = PTR_ERR(res); switch (error) { /* Make a negative dentry */ case -ENOENT: - inode = NULL; - break; + res = NULL; + goto out; /* This turned out not to be a regular file */ + case -EISDIR: + case -ENOTDIR: + goto no_open; case -ELOOP: if (!(nd->intent.open.flags & O_NOFOLLOW)) goto no_open; - /* case -EISDIR: */ /* case -EINVAL: */ default: goto out; } - } -no_entry: - d_add(dentry, inode); + } else if (res != NULL) + dentry = res; nfs_renew_times(dentry); nfs_set_verifier(dentry, nfs_save_change_attribute(dir)); out: - BUG_ON(error > 0); - return ERR_PTR(error); + return res; no_open: return nfs_lookup(dir, dentry, nd); } @@ -850,33 +1060,33 @@ static int nfs_open_revalidate(struct dentry *dentry, struct nameidata *nd) unsigned long verifier; int openflags, ret = 0; - /* NFS only supports OPEN for regular files */ - if (inode && !S_ISREG(inode->i_mode)) - goto no_open; parent = dget_parent(dentry); dir = parent->d_inode; if (!is_atomic_open(dir, nd)) goto no_open; + /* We can't create new files in nfs_open_revalidate(), so we + * optimize away revalidation of negative dentries. + */ + if (inode == NULL) + goto out; + /* NFS only supports OPEN on regular files */ + if (!S_ISREG(inode->i_mode)) + goto no_open; openflags = nd->intent.open.flags; - if (openflags & O_CREAT) { - /* If this is a negative dentry, just drop it */ - if (!inode) - goto out; - /* If this is exclusive open, just revalidate */ - if (openflags & O_EXCL) - goto no_open; - } + /* We cannot do exclusive creation on a positive dentry */ + if ((openflags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL)) + goto no_open; /* We can't create new files, or truncate existing ones here */ openflags &= ~(O_CREAT|O_TRUNC); /* - * Note: we're not holding inode->i_sem and so may be racing with + * Note: we're not holding inode->i_mutex and so may be racing with * operations that change the directory. We therefore save the * change attribute *before* we do the RPC call. */ lock_kernel(); verifier = nfs_save_change_attribute(dir); - ret = nfs4_open_revalidate(dir, dentry, openflags); + ret = nfs4_open_revalidate(dir, dentry, openflags, nd); if (!ret) nfs_set_verifier(dentry, verifier); unlock_kernel(); @@ -887,93 +1097,67 @@ out: return ret; no_open: dput(parent); + if (inode != NULL && nfs_have_delegation(inode, FMODE_READ)) + return 1; return nfs_lookup_revalidate(dentry, nd); } #endif /* CONFIG_NFSV4 */ -static inline -int find_dirent_name(nfs_readdir_descriptor_t *desc, struct page *page, struct dentry *dentry) +static struct dentry *nfs_readdir_lookup(nfs_readdir_descriptor_t *desc) { + struct dentry *parent = desc->file->f_path.dentry; + struct inode *dir = parent->d_inode; struct nfs_entry *entry = desc->entry; - int status; + struct dentry *dentry, *alias; + struct qstr name = { + .name = entry->name, + .len = entry->len, + }; + struct inode *inode; - while((status = dir_decode(desc)) == 0) { - if (entry->len != dentry->d_name.len) - continue; - if (memcmp(entry->name, dentry->d_name.name, entry->len)) - continue; - if (!(entry->fattr->valid & NFS_ATTR_FATTR)) - continue; - break; + switch (name.len) { + case 2: + if (name.name[0] == '.' && name.name[1] == '.') + return dget_parent(parent); + break; + case 1: + if (name.name[0] == '.') + return dget(parent); + } + name.hash = full_name_hash(name.name, name.len); + dentry = d_lookup(parent, &name); + if (dentry != NULL) + return dentry; + if (!desc->plus || !(entry->fattr->valid & NFS_ATTR_FATTR)) + return NULL; + /* Note: caller is already holding the dir->i_mutex! */ + dentry = d_alloc(parent, &name); + if (dentry == NULL) + return NULL; + dentry->d_op = NFS_PROTO(dir)->dentry_ops; + inode = nfs_fhget(dentry->d_sb, entry->fh, entry->fattr); + if (IS_ERR(inode)) { + dput(dentry); + return NULL; } - return status; -} - -/* - * Use the cached Readdirplus results in order to avoid a LOOKUP call - * whenever we believe that the parent directory has not changed. - * - * We assume that any file creation/rename changes the directory mtime. - * As this results in a page cache invalidation whenever it occurs, - * we don't require any other tests for cache coherency. - */ -static -int nfs_cached_lookup(struct inode *dir, struct dentry *dentry, - struct nfs_fh *fh, struct nfs_fattr *fattr) -{ - nfs_readdir_descriptor_t desc; - struct nfs_server *server; - struct nfs_entry entry; - struct page *page; - unsigned long timestamp; - int res; - - if (!NFS_USE_READDIRPLUS(dir)) - return -ENOENT; - server = NFS_SERVER(dir); - /* Don't use readdirplus unless the cache is stable */ - if ((server->flags & NFS_MOUNT_NOAC) != 0 - || nfs_caches_unstable(dir) - || nfs_attribute_timeout(dir)) - return -ENOENT; - if ((NFS_FLAGS(dir) & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA)) != 0) - return -ENOENT; - timestamp = NFS_I(dir)->readdir_timestamp; - - entry.fh = fh; - entry.fattr = fattr; - - desc.decode = NFS_PROTO(dir)->decode_dirent; - desc.entry = &entry; - desc.page_index = 0; - desc.plus = 1; - - for(;(page = find_get_page(dir->i_mapping, desc.page_index)); desc.page_index++) { - - res = -EIO; - if (PageUptodate(page)) { - void * kaddr = kmap_atomic(page, KM_USER0); - desc.ptr = kaddr; - res = find_dirent_name(&desc, page, dentry); - kunmap_atomic(kaddr, KM_USER0); - } - page_cache_release(page); - if (res == 0) - goto out_found; - if (res != -EAGAIN) - break; + alias = d_materialise_unique(dentry, inode); + if (alias != NULL) { + dput(dentry); + if (IS_ERR(alias)) + return NULL; + dentry = alias; } - return -ENOENT; - out_found: - fattr->timestamp = timestamp; - return 0; + + nfs_renew_times(dentry); + nfs_set_verifier(dentry, nfs_save_change_attribute(dir)); + return dentry; } /* * Code common to create, mkdir, and mknod. */ -static int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle, +int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle, struct nfs_fattr *fattr) { struct inode *inode; @@ -982,23 +1166,26 @@ static int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle, /* We may have been initialized further down */ if (dentry->d_inode) return 0; - if (fhandle->size == 0 || !(fattr->valid & NFS_ATTR_FATTR)) { + if (fhandle->size == 0) { struct inode *dir = dentry->d_parent->d_inode; error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr); if (error) - goto out_err; + return error; } - inode = nfs_fhget(dentry->d_sb, fhandle, fattr); - if (inode) { - d_instantiate(dentry, inode); - nfs_renew_times(dentry); - nfs_set_verifier(dentry, nfs_save_change_attribute(dentry->d_parent->d_inode)); - return 0; + if (!(fattr->valid & NFS_ATTR_FATTR)) { + struct nfs_server *server = NFS_SB(dentry->d_sb); + error = server->nfs_client->rpc_ops->getattr(server, fhandle, fattr); + if (error < 0) + return error; } - error = -ENOMEM; -out_err: - d_drop(dentry); - return error; + inode = nfs_fhget(dentry->d_sb, fhandle, fattr); + error = PTR_ERR(inode); + if (IS_ERR(inode)) + return error; + d_instantiate(dentry, inode); + if (d_unhashed(dentry)) + d_rehash(dentry); + return 0; } /* @@ -1011,12 +1198,11 @@ static int nfs_create(struct inode *dir, struct dentry *dentry, int mode, struct nameidata *nd) { struct iattr attr; - struct inode *inode; int error; int open_flags = 0; - dfprintk(VFS, "NFS: create(%s/%ld, %s\n", dir->i_sb->s_id, - dir->i_ino, dentry->d_name.name); + dfprintk(VFS, "NFS: create(%s/%ld), %s\n", + dir->i_sb->s_id, dir->i_ino, dentry->d_name.name); attr.ia_mode = mode; attr.ia_valid = ATTR_MODE; @@ -1024,26 +1210,19 @@ static int nfs_create(struct inode *dir, struct dentry *dentry, int mode, if (nd && (nd->flags & LOOKUP_CREATE)) open_flags = nd->intent.open.flags; - /* - * The 0 argument passed into the create function should one day - * contain the O_EXCL flag if requested. This allows NFSv3 to - * select the appropriate create strategy. Currently open_namei - * does not pass the create flags. - */ lock_kernel(); nfs_begin_data_update(dir); - inode = NFS_PROTO(dir)->create(dir, &dentry->d_name, &attr, open_flags); + error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, nd); nfs_end_data_update(dir); - if (!IS_ERR(inode)) { - d_instantiate(dentry, inode); - nfs_renew_times(dentry); - nfs_set_verifier(dentry, nfs_save_change_attribute(dir)); - error = 0; - } else { - error = PTR_ERR(inode); - d_drop(dentry); - } + if (error != 0) + goto out_err; + nfs_renew_times(dentry); + nfs_set_verifier(dentry, nfs_save_change_attribute(dir)); + unlock_kernel(); + return 0; +out_err: unlock_kernel(); + d_drop(dentry); return error; } @@ -1054,12 +1233,10 @@ static int nfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t rdev) { struct iattr attr; - struct nfs_fattr fattr; - struct nfs_fh fhandle; - int error; + int status; - dfprintk(VFS, "NFS: mknod(%s/%ld, %s\n", dir->i_sb->s_id, - dir->i_ino, dentry->d_name.name); + dfprintk(VFS, "NFS: mknod(%s/%ld), %s\n", + dir->i_sb->s_id, dir->i_ino, dentry->d_name.name); if (!new_valid_dev(rdev)) return -EINVAL; @@ -1069,15 +1246,18 @@ nfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t rdev) lock_kernel(); nfs_begin_data_update(dir); - error = NFS_PROTO(dir)->mknod(dir, &dentry->d_name, &attr, rdev, - &fhandle, &fattr); + status = NFS_PROTO(dir)->mknod(dir, dentry, &attr, rdev); nfs_end_data_update(dir); - if (!error) - error = nfs_instantiate(dentry, &fhandle, &fattr); - else - d_drop(dentry); + if (status != 0) + goto out_err; + nfs_renew_times(dentry); + nfs_set_verifier(dentry, nfs_save_change_attribute(dir)); unlock_kernel(); - return error; + return 0; +out_err: + unlock_kernel(); + d_drop(dentry); + return status; } /* @@ -1086,34 +1266,26 @@ nfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t rdev) static int nfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) { struct iattr attr; - struct nfs_fattr fattr; - struct nfs_fh fhandle; int error; - dfprintk(VFS, "NFS: mkdir(%s/%ld, %s\n", dir->i_sb->s_id, - dir->i_ino, dentry->d_name.name); + dfprintk(VFS, "NFS: mkdir(%s/%ld), %s\n", + dir->i_sb->s_id, dir->i_ino, dentry->d_name.name); attr.ia_valid = ATTR_MODE; attr.ia_mode = mode | S_IFDIR; lock_kernel(); -#if 0 - /* - * Always drop the dentry, we can't always depend on - * the fattr returned by the server (AIX seems to be - * broken). We're better off doing another lookup than - * depending on potentially bogus information. - */ - d_drop(dentry); -#endif nfs_begin_data_update(dir); - error = NFS_PROTO(dir)->mkdir(dir, &dentry->d_name, &attr, &fhandle, - &fattr); + error = NFS_PROTO(dir)->mkdir(dir, dentry, &attr); nfs_end_data_update(dir); - if (!error) - error = nfs_instantiate(dentry, &fhandle, &fattr); - else - d_drop(dentry); + if (error != 0) + goto out_err; + nfs_renew_times(dentry); + nfs_set_verifier(dentry, nfs_save_change_attribute(dir)); + unlock_kernel(); + return 0; +out_err: + d_drop(dentry); unlock_kernel(); return error; } @@ -1122,15 +1294,15 @@ static int nfs_rmdir(struct inode *dir, struct dentry *dentry) { int error; - dfprintk(VFS, "NFS: rmdir(%s/%ld, %s\n", dir->i_sb->s_id, - dir->i_ino, dentry->d_name.name); + dfprintk(VFS, "NFS: rmdir(%s/%ld), %s\n", + dir->i_sb->s_id, dir->i_ino, dentry->d_name.name); lock_kernel(); nfs_begin_data_update(dir); error = NFS_PROTO(dir)->rmdir(dir, &dentry->d_name); /* Ensure the VFS deletes this inode */ if (error == 0 && dentry->d_inode != NULL) - dentry->d_inode->i_nlink = 0; + clear_nlink(dentry->d_inode); nfs_end_data_update(dir); unlock_kernel(); @@ -1142,7 +1314,7 @@ static int nfs_sillyrename(struct inode *dir, struct dentry *dentry) static unsigned int sillycounter; const int i_inosize = sizeof(dir->i_ino)*2; const int countersize = sizeof(sillycounter)*2; - const int slen = strlen(".nfs") + i_inosize + countersize; + const int slen = sizeof(".nfs") + i_inosize + countersize - 1; char silly[slen+1]; struct qstr qsilly; struct dentry *sdentry; @@ -1151,6 +1323,7 @@ static int nfs_sillyrename(struct inode *dir, struct dentry *dentry) dfprintk(VFS, "NFS: silly-rename(%s/%s, ct=%d)\n", dentry->d_parent->d_name.name, dentry->d_name.name, atomic_read(&dentry->d_count)); + nfs_inc_stats(dir, NFSIOS_SILLYRENAME); #ifdef NFS_PARANOIA if (!dentry->d_inode) @@ -1167,6 +1340,9 @@ dentry->d_parent->d_name.name, dentry->d_name.name); sprintf(silly, ".nfs%*.*lx", i_inosize, i_inosize, dentry->d_inode->i_ino); + /* Return delegation in anticipation of the rename */ + nfs_inode_return_delegation(dentry->d_inode); + sdentry = NULL; do { char *suffix = silly + slen - countersize; @@ -1175,8 +1351,8 @@ dentry->d_parent->d_name.name, dentry->d_name.name); sillycounter++; sprintf(suffix, "%*.*x", countersize, countersize, sillycounter); - dfprintk(VFS, "trying to rename %s to %s\n", - dentry->d_name.name, silly); + dfprintk(VFS, "NFS: trying to rename %s to %s\n", + dentry->d_name.name, silly); sdentry = lookup_one_len(silly, dentry->d_parent, slen); /* @@ -1194,6 +1370,7 @@ dentry->d_parent->d_name.name, dentry->d_name.name); nfs_begin_data_update(dentry->d_inode); error = NFS_PROTO(dir)->rename(dir, &dentry->d_name, dir, &qsilly); + nfs_mark_for_revalidate(dentry->d_inode); nfs_end_data_update(dentry->d_inode); } else error = NFS_PROTO(dir)->rename(dir, &dentry->d_name, @@ -1235,11 +1412,13 @@ static int nfs_safe_remove(struct dentry *dentry) nfs_begin_data_update(dir); if (inode != NULL) { + nfs_inode_return_delegation(inode); nfs_begin_data_update(inode); error = NFS_PROTO(dir)->remove(dir, &dentry->d_name); /* The VFS may want to delete this inode */ if (error == 0) - inode->i_nlink--; + drop_nlink(inode); + nfs_mark_for_revalidate(inode); nfs_end_data_update(inode); } else error = NFS_PROTO(dir)->remove(dir, &dentry->d_name); @@ -1287,63 +1466,82 @@ static int nfs_unlink(struct inode *dir, struct dentry *dentry) return error; } -static int -nfs_symlink(struct inode *dir, struct dentry *dentry, const char *symname) +/* + * To create a symbolic link, most file systems instantiate a new inode, + * add a page to it containing the path, then write it out to the disk + * using prepare_write/commit_write. + * + * Unfortunately the NFS client can't create the in-core inode first + * because it needs a file handle to create an in-core inode (see + * fs/nfs/inode.c:nfs_fhget). We only have a file handle *after* the + * symlink request has completed on the server. + * + * So instead we allocate a raw page, copy the symname into it, then do + * the SYMLINK request with the page as the buffer. If it succeeds, we + * now have a new file handle and can instantiate an in-core NFS inode + * and move the raw page into its mapping. + */ +static int nfs_symlink(struct inode *dir, struct dentry *dentry, const char *symname) { + struct pagevec lru_pvec; + struct page *page; + char *kaddr; struct iattr attr; - struct nfs_fattr sym_attr; - struct nfs_fh sym_fh; - struct qstr qsymname; + unsigned int pathlen = strlen(symname); int error; dfprintk(VFS, "NFS: symlink(%s/%ld, %s, %s)\n", dir->i_sb->s_id, dir->i_ino, dentry->d_name.name, symname); - error = -ENAMETOOLONG; - switch (NFS_PROTO(dir)->version) { - case 2: - if (strlen(symname) > NFS2_MAXPATHLEN) - goto out; - break; - case 3: - if (strlen(symname) > NFS3_MAXPATHLEN) - goto out; - default: - break; - } + if (pathlen > PAGE_SIZE) + return -ENAMETOOLONG; -#ifdef NFS_PARANOIA -if (dentry->d_inode) -printk("nfs_proc_symlink: %s/%s not negative!\n", -dentry->d_parent->d_name.name, dentry->d_name.name); -#endif - /* - * Fill in the sattr for the call. - * Note: SunOS 4.1.2 crashes if the mode isn't initialized! - */ - attr.ia_valid = ATTR_MODE; attr.ia_mode = S_IFLNK | S_IRWXUGO; - - qsymname.name = symname; - qsymname.len = strlen(symname); + attr.ia_valid = ATTR_MODE; lock_kernel(); + + page = alloc_page(GFP_KERNEL); + if (!page) { + unlock_kernel(); + return -ENOMEM; + } + + kaddr = kmap_atomic(page, KM_USER0); + memcpy(kaddr, symname, pathlen); + if (pathlen < PAGE_SIZE) + memset(kaddr + pathlen, 0, PAGE_SIZE - pathlen); + kunmap_atomic(kaddr, KM_USER0); + nfs_begin_data_update(dir); - error = NFS_PROTO(dir)->symlink(dir, &dentry->d_name, &qsymname, - &attr, &sym_fh, &sym_attr); + error = NFS_PROTO(dir)->symlink(dir, dentry, page, pathlen, &attr); nfs_end_data_update(dir); - if (!error) { - error = nfs_instantiate(dentry, &sym_fh, &sym_attr); - } else { - if (error == -EEXIST) - printk("nfs_proc_symlink: %s/%s already exists??\n", - dentry->d_parent->d_name.name, dentry->d_name.name); + if (error != 0) { + dfprintk(VFS, "NFS: symlink(%s/%ld, %s, %s) error %d\n", + dir->i_sb->s_id, dir->i_ino, + dentry->d_name.name, symname, error); d_drop(dentry); + __free_page(page); + unlock_kernel(); + return error; } - unlock_kernel(); -out: - return error; + /* + * No big deal if we can't add this page to the page cache here. + * READLINK will get the missing page from the server if needed. + */ + pagevec_init(&lru_pvec, 0); + if (!add_to_page_cache(page, dentry->d_inode->i_mapping, 0, + GFP_KERNEL)) { + pagevec_add(&lru_pvec, page); + pagevec_lru_add(&lru_pvec); + SetPageUptodate(page); + unlock_page(page); + } else + __free_page(page); + + unlock_kernel(); + return 0; } static int @@ -1356,17 +1554,14 @@ nfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry) old_dentry->d_parent->d_name.name, old_dentry->d_name.name, dentry->d_parent->d_name.name, dentry->d_name.name); - /* - * Drop the dentry in advance to force a new lookup. - * Since nfs_proc_link doesn't return a file handle, - * we can't use the existing dentry. - */ lock_kernel(); - d_drop(dentry); - nfs_begin_data_update(dir); nfs_begin_data_update(inode); error = NFS_PROTO(dir)->link(inode, dir, &dentry->d_name); + if (error == 0) { + atomic_inc(&inode->i_count); + d_instantiate(dentry, inode); + } nfs_end_data_update(inode); nfs_end_data_update(dir); unlock_kernel(); @@ -1430,9 +1625,11 @@ static int nfs_rename(struct inode *old_dir, struct dentry *old_dentry, */ if (!new_inode) goto go_ahead; - if (S_ISDIR(new_inode->i_mode)) - goto out; - else if (atomic_read(&new_dentry->d_count) > 1) { + if (S_ISDIR(new_inode->i_mode)) { + error = -EISDIR; + if (!S_ISDIR(old_inode->i_mode)) + goto out; + } else if (atomic_read(&new_dentry->d_count) > 2) { int err; /* copy the target dentry's name */ dentry = d_alloc(new_dentry->d_parent, @@ -1447,10 +1644,8 @@ static int nfs_rename(struct inode *old_dir, struct dentry *old_dentry, new_inode = NULL; /* instantiate the replacement target */ d_instantiate(new_dentry, NULL); - } - + } else if (atomic_read(&new_dentry->d_count) > 1) { /* dentry still busy? */ - if (atomic_read(&new_dentry->d_count) > 1) { #ifdef NFS_PARANOIA printk("nfs_rename: target %s/%s busy, d_count=%d\n", new_dentry->d_parent->d_name.name, @@ -1459,25 +1654,31 @@ static int nfs_rename(struct inode *old_dir, struct dentry *old_dentry, #endif goto out; } - } + } else + drop_nlink(new_inode); go_ahead: /* * ... prune child dentries and writebacks if needed. */ if (atomic_read(&old_dentry->d_count) > 1) { - nfs_wb_all(old_inode); + if (S_ISREG(old_inode->i_mode)) + nfs_wb_all(old_inode); shrink_dcache_parent(old_dentry); } + nfs_inode_return_delegation(old_inode); - if (new_inode) + if (new_inode != NULL) { + nfs_inode_return_delegation(new_inode); d_delete(new_dentry); + } nfs_begin_data_update(old_dir); nfs_begin_data_update(new_dir); nfs_begin_data_update(old_inode); error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name, new_dir, &new_dentry->d_name); + nfs_mark_for_revalidate(old_inode); nfs_end_data_update(old_inode); nfs_end_data_update(new_dir); nfs_end_data_update(old_dir); @@ -1485,8 +1686,7 @@ out: if (rehash) d_rehash(rehash); if (!error) { - if (!S_ISDIR(old_inode->i_mode)) - d_move(old_dentry, new_dentry); + d_move(old_dentry, new_dentry); nfs_renew_times(new_dentry); nfs_set_verifier(new_dentry, nfs_save_change_attribute(new_dir)); } @@ -1498,85 +1698,291 @@ out: return error; } -int -nfs_permission(struct inode *inode, int mask, struct nameidata *nd) +static DEFINE_SPINLOCK(nfs_access_lru_lock); +static LIST_HEAD(nfs_access_lru_list); +static atomic_long_t nfs_access_nr_entries; + +static void nfs_access_free_entry(struct nfs_access_entry *entry) +{ + put_rpccred(entry->cred); + kfree(entry); + smp_mb__before_atomic_dec(); + atomic_long_dec(&nfs_access_nr_entries); + smp_mb__after_atomic_dec(); +} + +int nfs_access_cache_shrinker(int nr_to_scan, gfp_t gfp_mask) +{ + LIST_HEAD(head); + struct nfs_inode *nfsi; + struct nfs_access_entry *cache; + + spin_lock(&nfs_access_lru_lock); +restart: + list_for_each_entry(nfsi, &nfs_access_lru_list, access_cache_inode_lru) { + struct inode *inode; + + if (nr_to_scan-- == 0) + break; + inode = igrab(&nfsi->vfs_inode); + if (inode == NULL) + continue; + spin_lock(&inode->i_lock); + if (list_empty(&nfsi->access_cache_entry_lru)) + goto remove_lru_entry; + cache = list_entry(nfsi->access_cache_entry_lru.next, + struct nfs_access_entry, lru); + list_move(&cache->lru, &head); + rb_erase(&cache->rb_node, &nfsi->access_cache); + if (!list_empty(&nfsi->access_cache_entry_lru)) + list_move_tail(&nfsi->access_cache_inode_lru, + &nfs_access_lru_list); + else { +remove_lru_entry: + list_del_init(&nfsi->access_cache_inode_lru); + clear_bit(NFS_INO_ACL_LRU_SET, &nfsi->flags); + } + spin_unlock(&inode->i_lock); + iput(inode); + goto restart; + } + spin_unlock(&nfs_access_lru_lock); + while (!list_empty(&head)) { + cache = list_entry(head.next, struct nfs_access_entry, lru); + list_del(&cache->lru); + nfs_access_free_entry(cache); + } + return (atomic_long_read(&nfs_access_nr_entries) / 100) * sysctl_vfs_cache_pressure; +} + +static void __nfs_access_zap_cache(struct inode *inode) +{ + struct nfs_inode *nfsi = NFS_I(inode); + struct rb_root *root_node = &nfsi->access_cache; + struct rb_node *n, *dispose = NULL; + struct nfs_access_entry *entry; + + /* Unhook entries from the cache */ + while ((n = rb_first(root_node)) != NULL) { + entry = rb_entry(n, struct nfs_access_entry, rb_node); + rb_erase(n, root_node); + list_del(&entry->lru); + n->rb_left = dispose; + dispose = n; + } + nfsi->cache_validity &= ~NFS_INO_INVALID_ACCESS; + spin_unlock(&inode->i_lock); + + /* Now kill them all! */ + while (dispose != NULL) { + n = dispose; + dispose = n->rb_left; + nfs_access_free_entry(rb_entry(n, struct nfs_access_entry, rb_node)); + } +} + +void nfs_access_zap_cache(struct inode *inode) +{ + /* Remove from global LRU init */ + if (test_and_clear_bit(NFS_INO_ACL_LRU_SET, &NFS_FLAGS(inode))) { + spin_lock(&nfs_access_lru_lock); + list_del_init(&NFS_I(inode)->access_cache_inode_lru); + spin_unlock(&nfs_access_lru_lock); + } + + spin_lock(&inode->i_lock); + /* This will release the spinlock */ + __nfs_access_zap_cache(inode); +} + +static struct nfs_access_entry *nfs_access_search_rbtree(struct inode *inode, struct rpc_cred *cred) +{ + struct rb_node *n = NFS_I(inode)->access_cache.rb_node; + struct nfs_access_entry *entry; + + while (n != NULL) { + entry = rb_entry(n, struct nfs_access_entry, rb_node); + + if (cred < entry->cred) + n = n->rb_left; + else if (cred > entry->cred) + n = n->rb_right; + else + return entry; + } + return NULL; +} + +int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, struct nfs_access_entry *res) +{ + struct nfs_inode *nfsi = NFS_I(inode); + struct nfs_access_entry *cache; + int err = -ENOENT; + + spin_lock(&inode->i_lock); + if (nfsi->cache_validity & NFS_INO_INVALID_ACCESS) + goto out_zap; + cache = nfs_access_search_rbtree(inode, cred); + if (cache == NULL) + goto out; + if (time_after(jiffies, cache->jiffies + NFS_ATTRTIMEO(inode))) + goto out_stale; + res->jiffies = cache->jiffies; + res->cred = cache->cred; + res->mask = cache->mask; + list_move_tail(&cache->lru, &nfsi->access_cache_entry_lru); + err = 0; +out: + spin_unlock(&inode->i_lock); + return err; +out_stale: + rb_erase(&cache->rb_node, &nfsi->access_cache); + list_del(&cache->lru); + spin_unlock(&inode->i_lock); + nfs_access_free_entry(cache); + return -ENOENT; +out_zap: + /* This will release the spinlock */ + __nfs_access_zap_cache(inode); + return -ENOENT; +} + +static void nfs_access_add_rbtree(struct inode *inode, struct nfs_access_entry *set) +{ + struct nfs_inode *nfsi = NFS_I(inode); + struct rb_root *root_node = &nfsi->access_cache; + struct rb_node **p = &root_node->rb_node; + struct rb_node *parent = NULL; + struct nfs_access_entry *entry; + + spin_lock(&inode->i_lock); + while (*p != NULL) { + parent = *p; + entry = rb_entry(parent, struct nfs_access_entry, rb_node); + + if (set->cred < entry->cred) + p = &parent->rb_left; + else if (set->cred > entry->cred) + p = &parent->rb_right; + else + goto found; + } + rb_link_node(&set->rb_node, parent, p); + rb_insert_color(&set->rb_node, root_node); + list_add_tail(&set->lru, &nfsi->access_cache_entry_lru); + spin_unlock(&inode->i_lock); + return; +found: + rb_replace_node(parent, &set->rb_node, root_node); + list_add_tail(&set->lru, &nfsi->access_cache_entry_lru); + list_del(&entry->lru); + spin_unlock(&inode->i_lock); + nfs_access_free_entry(entry); +} + +void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set) +{ + struct nfs_access_entry *cache = kmalloc(sizeof(*cache), GFP_KERNEL); + if (cache == NULL) + return; + RB_CLEAR_NODE(&cache->rb_node); + cache->jiffies = set->jiffies; + cache->cred = get_rpccred(set->cred); + cache->mask = set->mask; + + nfs_access_add_rbtree(inode, cache); + + /* Update accounting */ + smp_mb__before_atomic_inc(); + atomic_long_inc(&nfs_access_nr_entries); + smp_mb__after_atomic_inc(); + + /* Add inode to global LRU list */ + if (!test_and_set_bit(NFS_INO_ACL_LRU_SET, &NFS_FLAGS(inode))) { + spin_lock(&nfs_access_lru_lock); + list_add_tail(&NFS_I(inode)->access_cache_inode_lru, &nfs_access_lru_list); + spin_unlock(&nfs_access_lru_lock); + } +} + +static int nfs_do_access(struct inode *inode, struct rpc_cred *cred, int mask) +{ + struct nfs_access_entry cache; + int status; + + status = nfs_access_get_cached(inode, cred, &cache); + if (status == 0) + goto out; + + /* Be clever: ask server to check for all possible rights */ + cache.mask = MAY_EXEC | MAY_WRITE | MAY_READ; + cache.cred = cred; + cache.jiffies = jiffies; + status = NFS_PROTO(inode)->access(inode, &cache); + if (status != 0) + return status; + nfs_access_add_cache(inode, &cache); +out: + if ((cache.mask & mask) == mask) + return 0; + return -EACCES; +} + +int nfs_permission(struct inode *inode, int mask, struct nameidata *nd) { - struct nfs_access_cache *cache = &NFS_I(inode)->cache_access; struct rpc_cred *cred; - int mode = inode->i_mode; - int res; + int res = 0; + + nfs_inc_stats(inode, NFSIOS_VFSACCESS); if (mask == 0) - return 0; - if (mask & MAY_WRITE) { - /* - * - * Nobody gets write access to a read-only fs. - * - */ - if (IS_RDONLY(inode) && - (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) - return -EROFS; + goto out; + /* Is this sys_access() ? */ + if (nd != NULL && (nd->flags & LOOKUP_ACCESS)) + goto force_lookup; - /* - * - * Nobody gets write access to an immutable file. - * - */ - if (IS_IMMUTABLE(inode)) - return -EACCES; - } - /* Are we checking permissions on anything other than lookup/execute? */ - if ((mask & MAY_EXEC) == 0) { - /* We only need to check permissions on file open() and access() */ - if (!nd || !(nd->flags & (LOOKUP_OPEN|LOOKUP_ACCESS))) - return 0; - /* NFSv4 has atomic_open... */ - if (NFS_PROTO(inode)->version > 3 && (nd->flags & LOOKUP_OPEN)) - return 0; + switch (inode->i_mode & S_IFMT) { + case S_IFLNK: + goto out; + case S_IFREG: + /* NFSv4 has atomic_open... */ + if (nfs_server_capable(inode, NFS_CAP_ATOMIC_OPEN) + && nd != NULL + && (nd->flags & LOOKUP_OPEN)) + goto out; + break; + case S_IFDIR: + /* + * Optimize away all write operations, since the server + * will check permissions when we perform the op. + */ + if ((mask & MAY_WRITE) && !(mask & MAY_READ)) + goto out; } +force_lookup: lock_kernel(); if (!NFS_PROTO(inode)->access) goto out_notsup; cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0); - if (cache->cred == cred - && time_before(jiffies, cache->jiffies + NFS_ATTRTIMEO(inode)) - && !(NFS_FLAGS(inode) & NFS_INO_INVALID_ATTR)) { - if (!(res = cache->err)) { - /* Is the mask a subset of an accepted mask? */ - if ((cache->mask & mask) == mask) - goto out; - } else { - /* ...or is it a superset of a rejected mask? */ - if ((cache->mask & mask) == cache->mask) - goto out; - } - } - - res = NFS_PROTO(inode)->access(inode, cred, mask); - if (!res || res == -EACCES) - goto add_cache; -out: - put_rpccred(cred); + if (!IS_ERR(cred)) { + res = nfs_do_access(inode, cred, mask); + put_rpccred(cred); + } else + res = PTR_ERR(cred); unlock_kernel(); +out: + dfprintk(VFS, "NFS: permission(%s/%ld), mask=0x%x, res=%d\n", + inode->i_sb->s_id, inode->i_ino, mask, res); return res; out_notsup: - nfs_revalidate_inode(NFS_SERVER(inode), inode); - res = vfs_permission(inode, mask); - unlock_kernel(); - return res; -add_cache: - cache->jiffies = jiffies; - if (cache->cred) - put_rpccred(cache->cred); - cache->cred = cred; - cache->mask = mask; - cache->err = res; + res = nfs_revalidate_inode(NFS_SERVER(inode), inode); + if (res == 0) + res = generic_permission(inode, mask, NULL); unlock_kernel(); - return res; + goto out; } /*