struct dentry *
cifs_lookup(struct inode *parent_dir_inode, struct dentry *direntry,
struct nameidata *nd)
{
int xid;
int rc = 0; /* to get around spurious gcc warning, set to zero here */
__u32 oplock = 0;
__u16 fileHandle = 0;
bool posix_open = false;
struct cifs_sb_info *cifs_sb;
struct cifsTconInfo *pTcon;
struct cifsFileInfo *cfile;
struct inode *newInode = NULL;
char *full_path = NULL;
struct file *filp;
xid = GetXid();
cFYI(1, "parent inode = 0x%p name is: %s and dentry = 0x%p",
parent_dir_inode, direntry->d_name.name, direntry);
/* check whether path exists */
cifs_sb = CIFS_SB(parent_dir_inode->i_sb);
pTcon = cifs_sb->tcon;
/*
* Don't allow the separator character in a path component.
* The VFS will not allow "/", but "\" is allowed by posix.
*/
if (!(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_POSIX_PATHS)) {
int i;
for (i = 0; i < direntry->d_name.len; i++)
if (direntry->d_name.name[i] == '\\') {
cFYI(1, "Invalid file name");
FreeXid(xid);
return ERR_PTR(-EINVAL);
}
}
/*
* O_EXCL: optimize away the lookup, but don't hash the dentry. Let
* the VFS handle the create.
*/
if (nd && (nd->flags & LOOKUP_EXCL)) {
d_instantiate(direntry, NULL);
return NULL;
}
/* can not grab the rename sem here since it would
deadlock in the cases (beginning of sys_rename itself)
in which we already have the sb rename sem */
full_path = build_path_from_dentry(direntry);
if (full_path == NULL) {
FreeXid(xid);
return ERR_PTR(-ENOMEM);
}
if (direntry->d_inode != NULL) {
cFYI(1, "non-NULL inode in lookup");
} else {
cFYI(1, "NULL inode in lookup");
}
cFYI(1, "Full path: %s inode = 0x%p", full_path, direntry->d_inode);
/* Posix open is only called (at lookup time) for file create now.
* For opens (rather than creates), because we do not know if it
* is a file or directory yet, and current Samba no longer allows
* us to do posix open on dirs, we could end up wasting an open call
* on what turns out to be a dir. For file opens, we wait to call posix
* open till cifs_open. It could be added here (lookup) in the future
* but the performance tradeoff of the extra network request when EISDIR
* or EACCES is returned would have to be weighed against the 50%
* reduction in network traffic in the other paths.
*/
if (pTcon->unix_ext) {
if (nd && !(nd->flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY)) &&
(nd->flags & LOOKUP_OPEN) && !pTcon->broken_posix_open &&
(nd->intent.open.flags & O_CREAT)) {
rc = cifs_posix_open(full_path, &newInode,
parent_dir_inode->i_sb,
nd->intent.open.create_mode,
nd->intent.open.flags, &oplock,
&fileHandle, xid);
/*
* The check below works around a bug in POSIX
* open in samba versions 3.3.1 and earlier where
* open could incorrectly fail with invalid parameter.
* If either that or op not supported returned, follow
* the normal lookup.
*/
if ((rc == 0) || (rc == -ENOENT))
posix_open = true;
else if ((rc == -EINVAL) || (rc != -EOPNOTSUPP))
pTcon->broken_posix_open = true;
}
if (!posix_open)
rc = cifs_get_inode_info_unix(&newInode, full_path,
parent_dir_inode->i_sb, xid);
} else
rc = cifs_get_inode_info(&newInode, full_path, NULL,
parent_dir_inode->i_sb, xid, NULL);
if ((rc == 0) && (newInode != NULL)) {
if (pTcon->nocase)
direntry->d_op = &cifs_ci_dentry_ops;
else
direntry->d_op = &cifs_dentry_ops;
d_add(direntry, newInode);
if (posix_open) {
filp = lookup_instantiate_filp(nd, direntry,
generic_file_open);
if (IS_ERR(filp)) {
rc = PTR_ERR(filp);
CIFSSMBClose(xid, pTcon, fileHandle);
goto lookup_out;
}
cfile = cifs_new_fileinfo(newInode, fileHandle, filp,
nd->path.mnt,
nd->intent.open.flags,
oplock);
if (cfile == NULL) {
fput(filp);
CIFSSMBClose(xid, pTcon, fileHandle);
rc = -ENOMEM;
goto lookup_out;
}
}
/* since paths are not looked up by component - the parent
directories are presumed to be good here */
renew_parental_timestamps(direntry);
} else if (rc == -ENOENT) {
rc = 0;
direntry->d_time = jiffies;
if (pTcon->nocase)
direntry->d_op = &cifs_ci_dentry_ops;
else
direntry->d_op = &cifs_dentry_ops;
d_add(direntry, NULL);
/* if it was once a directory (but how can we tell?) we could do
shrink_dcache_parent(direntry); */
} else if (rc != -EACCES) {
cERROR(1, "Unexpected lookup error %d", rc);
/* We special case check for Access Denied - since that
is a common return code */
}
lookup_out:
kfree(full_path);
FreeXid(xid);
return ERR_PTR(rc);
}
int
cifs_create(struct inode *inode, struct dentry *direntry, int mode,
struct nameidata *nd)
{
int rc = -ENOENT;
int xid;
int create_options = CREATE_NOT_DIR;
__u32 oplock = 0;
int oflags;
/*
* BB below access is probably too much for mknod to request
* but we have to do query and setpathinfo so requesting
* less could fail (unless we want to request getatr and setatr
* permissions (only). At least for POSIX we do not have to
* request so much.
*/
int desiredAccess = GENERIC_READ | GENERIC_WRITE;
__u16 fileHandle;
struct cifs_sb_info *cifs_sb;
struct cifsTconInfo *tcon;
char *full_path = NULL;
FILE_ALL_INFO *buf = NULL;
struct inode *newinode = NULL;
int disposition = FILE_OVERWRITE_IF;
xid = GetXid();
cifs_sb = CIFS_SB(inode->i_sb);
tcon = cifs_sb->tcon;
full_path = build_path_from_dentry(direntry);
if (full_path == NULL) {
rc = -ENOMEM;
goto cifs_create_out;
}
if (oplockEnabled)
oplock = REQ_OPLOCK;
if (nd && (nd->flags & LOOKUP_OPEN))
oflags = nd->intent.open.flags;
else
oflags = FMODE_READ | SMB_O_CREAT;
if (tcon->unix_ext && (tcon->ses->capabilities & CAP_UNIX) &&
(CIFS_UNIX_POSIX_PATH_OPS_CAP &
le64_to_cpu(tcon->fsUnixInfo.Capability))) {
rc = cifs_posix_open(full_path, &newinode,
inode->i_sb, mode, oflags, &oplock, &fileHandle, xid);
/* EIO could indicate that (posix open) operation is not
supported, despite what server claimed in capability
negotation. EREMOTE indicates DFS junction, which is not
handled in posix open */
if (rc == 0) {
if (newinode == NULL) /* query inode info */
goto cifs_create_get_file_info;
else /* success, no need to query */
goto cifs_create_set_dentry;
} else if ((rc != -EIO) && (rc != -EREMOTE) &&
(rc != -EOPNOTSUPP) && (rc != -EINVAL))
goto cifs_create_out;
/* else fallthrough to retry, using older open call, this is
case where server does not support this SMB level, and
falsely claims capability (also get here for DFS case
which should be rare for path not covered on files) */
}
if (nd && (nd->flags & LOOKUP_OPEN)) {
/* if the file is going to stay open, then we
need to set the desired access properly */
desiredAccess = 0;
if (oflags & FMODE_READ)
desiredAccess |= GENERIC_READ; /* is this too little? */
if (oflags & FMODE_WRITE)
desiredAccess |= GENERIC_WRITE;
if ((oflags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))
disposition = FILE_CREATE;
else if ((oflags & (O_CREAT | O_TRUNC)) == (O_CREAT | O_TRUNC))
disposition = FILE_OVERWRITE_IF;
else if ((oflags & O_CREAT) == O_CREAT)
disposition = FILE_OPEN_IF;
else
cFYI(1, "Create flag not set in create function");
}
/* BB add processing to set equivalent of mode - e.g. via CreateX with
ACLs */
buf = kmalloc(sizeof(FILE_ALL_INFO), GFP_KERNEL);
if (buf == NULL) {
rc = -ENOMEM;
goto cifs_create_out;
}
/*
* if we're not using unix extensions, see if we need to set
* ATTR_READONLY on the create call
*/
if (!tcon->unix_ext && (mode & S_IWUGO) == 0)
create_options |= CREATE_OPTION_READONLY;
if (cifs_sb->tcon->ses->capabilities & CAP_NT_SMBS)
rc = CIFSSMBOpen(xid, tcon, full_path, disposition,
desiredAccess, create_options,
&fileHandle, &oplock, buf, cifs_sb->local_nls,
cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MAP_SPECIAL_CHR);
else
rc = -EIO; /* no NT SMB support fall into legacy open below */
if (rc == -EIO) {
/* old server, retry the open legacy style */
rc = SMBLegacyOpen(xid, tcon, full_path, disposition,
desiredAccess, create_options,
&fileHandle, &oplock, buf, cifs_sb->local_nls,
cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MAP_SPECIAL_CHR);
}
if (rc) {
cFYI(1, "cifs_create returned 0x%x", rc);
goto cifs_create_out;
}
/* If Open reported that we actually created a file
then we now have to set the mode if possible */
if ((tcon->unix_ext) && (oplock & CIFS_CREATE_ACTION)) {
struct cifs_unix_set_info_args args = {
.mode = mode,
.ctime = NO_CHANGE_64,
.atime = NO_CHANGE_64,
.mtime = NO_CHANGE_64,
.device = 0,
};
if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SET_UID) {
args.uid = (__u64) current_fsuid();
if (inode->i_mode & S_ISGID)
args.gid = (__u64) inode->i_gid;
else
args.gid = (__u64) current_fsgid();
} else {
args.uid = NO_CHANGE_64;
args.gid = NO_CHANGE_64;
}
CIFSSMBUnixSetPathInfo(xid, tcon, full_path, &args,
cifs_sb->local_nls,
cifs_sb->mnt_cifs_flags &
CIFS_MOUNT_MAP_SPECIAL_CHR);
} else {
/* BB implement mode setting via Windows security
descriptors e.g. */
/* CIFSSMBWinSetPerms(xid,tcon,path,mode,-1,-1,nls);*/
/* Could set r/o dos attribute if mode & 0222 == 0 */
}
cifs_create_get_file_info:
/* server might mask mode so we have to query for it */
if (tcon->unix_ext)
rc = cifs_get_inode_info_unix(&newinode, full_path,
inode->i_sb, xid);
else {
rc = cifs_get_inode_info(&newinode, full_path, buf,
inode->i_sb, xid, &fileHandle);
if (newinode) {
if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_DYNPERM)
newinode->i_mode = mode;
if ((oplock & CIFS_CREATE_ACTION) &&
(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SET_UID)) {
newinode->i_uid = current_fsuid();
if (inode->i_mode & S_ISGID)
newinode->i_gid = inode->i_gid;
else
newinode->i_gid = current_fsgid();
}
}
}
cifs_create_set_dentry:
if (rc == 0)
setup_cifs_dentry(tcon, direntry, newinode);
else
cFYI(1, "Create worked, get_inode_info failed rc = %d", rc);
if (newinode && nd && (nd->flags & LOOKUP_OPEN)) {
struct cifsFileInfo *pfile_info;
struct file *filp;
filp = lookup_instantiate_filp(nd, direntry, generic_file_open);
if (IS_ERR(filp)) {
rc = PTR_ERR(filp);
CIFSSMBClose(xid, tcon, fileHandle);
goto cifs_create_out;
}
pfile_info = cifs_new_fileinfo(newinode, fileHandle, filp,
nd->path.mnt, oflags, oplock);
if (pfile_info == NULL) {
fput(filp);
CIFSSMBClose(xid, tcon, fileHandle);
rc = -ENOMEM;
}
} else {
CIFSSMBClose(xid, tcon, fileHandle);
}
cifs_create_out:
kfree(buf);
kfree(full_path);
FreeXid(xid);
return rc;
}
static int
v9fs_vfs_create_dotl(struct inode *dir, struct dentry *dentry, int omode,
struct nameidata *nd)
{
int err = 0;
gid_t gid;
int flags;
mode_t mode;
char *name = NULL;
struct file *filp;
struct p9_qid qid;
struct inode *inode;
struct p9_fid *fid = NULL;
struct v9fs_inode *v9inode;
struct p9_fid *dfid, *ofid, *inode_fid;
struct v9fs_session_info *v9ses;
struct posix_acl *pacl = NULL, *dacl = NULL;
v9ses = v9fs_inode2v9ses(dir);
if (nd && nd->flags & LOOKUP_OPEN)
flags = nd->intent.open.flags - 1;
else {
/*
* create call without LOOKUP_OPEN is due
* to mknod of regular files. So use mknod
* operation.
*/
return v9fs_vfs_mknod_dotl(dir, dentry, omode, 0);
}
name = (char *) dentry->d_name.name;
P9_DPRINTK(P9_DEBUG_VFS, "v9fs_vfs_create_dotl: name:%s flags:0x%x "
"mode:0x%x\n", name, flags, omode);
dfid = v9fs_fid_lookup(dentry->d_parent);
if (IS_ERR(dfid)) {
err = PTR_ERR(dfid);
P9_DPRINTK(P9_DEBUG_VFS, "fid lookup failed %d\n", err);
return err;
}
/* clone a fid to use for creation */
ofid = p9_client_walk(dfid, 0, NULL, 1);
if (IS_ERR(ofid)) {
err = PTR_ERR(ofid);
P9_DPRINTK(P9_DEBUG_VFS, "p9_client_walk failed %d\n", err);
return err;
}
gid = v9fs_get_fsgid_for_create(dir);
mode = omode;
/* Update mode based on ACL value */
err = v9fs_acl_mode(dir, &mode, &dacl, &pacl);
if (err) {
P9_DPRINTK(P9_DEBUG_VFS,
"Failed to get acl values in creat %d\n", err);
goto error;
}
err = p9_client_create_dotl(ofid, name, flags, mode, gid, &qid);
if (err < 0) {
P9_DPRINTK(P9_DEBUG_VFS,
"p9_client_open_dotl failed in creat %d\n",
err);
goto error;
}
v9fs_invalidate_inode_attr(dir);
/* instantiate inode and assign the unopened fid to the dentry */
fid = p9_client_walk(dfid, 1, &name, 1);
if (IS_ERR(fid)) {
err = PTR_ERR(fid);
P9_DPRINTK(P9_DEBUG_VFS, "p9_client_walk failed %d\n", err);
fid = NULL;
goto error;
}
inode = v9fs_get_inode_from_fid(v9ses, fid, dir->i_sb);
if (IS_ERR(inode)) {
err = PTR_ERR(inode);
P9_DPRINTK(P9_DEBUG_VFS, "inode creation failed %d\n", err);
goto error;
}
d_instantiate(dentry, inode);
err = v9fs_fid_add(dentry, fid);
if (err < 0)
goto error;
/* Now set the ACL based on the default value */
v9fs_set_create_acl(dentry, dacl, pacl);
v9inode = V9FS_I(inode);
mutex_lock(&v9inode->v_mutex);
if (v9ses->cache && !v9inode->writeback_fid &&
((flags & O_ACCMODE) != O_RDONLY)) {
/*
* clone a fid and add it to writeback_fid
* we do it during open time instead of
* page dirty time via write_begin/page_mkwrite
* because we want write after unlink usecase
* to work.
*/
inode_fid = v9fs_writeback_fid(dentry);
if (IS_ERR(inode_fid)) {
err = PTR_ERR(inode_fid);
mutex_unlock(&v9inode->v_mutex);
goto error;
}
v9inode->writeback_fid = (void *) inode_fid;
}
mutex_unlock(&v9inode->v_mutex);
/* Since we are opening a file, assign the open fid to the file */
filp = lookup_instantiate_filp(nd, dentry, generic_file_open);
if (IS_ERR(filp)) {
p9_client_clunk(ofid);
return PTR_ERR(filp);
}
filp->private_data = ofid;
#ifdef CONFIG_9P_FSCACHE
if (v9ses->cache)
v9fs_cache_inode_set_cookie(inode, filp);
#endif
return 0;
error:
if (ofid)
p9_client_clunk(ofid);
if (fid)
p9_client_clunk(fid);
return err;
}
/*
* Atomic create+open operation
*
* If the filesystem doesn't support this, then fall back to separate
* 'mknod' + 'open' requests.
*/
static int fuse_create_open(struct inode *dir, struct dentry *entry,
umode_t mode, struct nameidata *nd)
{
int err;
struct inode *inode;
struct fuse_conn *fc = get_fuse_conn(dir);
struct fuse_req *req;
struct fuse_forget_link *forget;
struct fuse_create_in inarg;
struct fuse_open_out outopen;
struct fuse_entry_out outentry;
struct fuse_file *ff;
struct file *file;
int flags = nd->intent.open.flags;
if (fc->no_create)
return -ENOSYS;
forget = fuse_alloc_forget();
if (!forget)
return -ENOMEM;
req = fuse_get_req(fc);
err = PTR_ERR(req);
if (IS_ERR(req))
goto out_put_forget_req;
err = -ENOMEM;
ff = fuse_file_alloc(fc);
if (!ff)
goto out_put_request;
if (!fc->dont_mask)
mode &= ~current_umask();
flags &= ~O_NOCTTY;
memset(&inarg, 0, sizeof(inarg));
memset(&outentry, 0, sizeof(outentry));
inarg.flags = flags;
inarg.mode = mode;
inarg.umask = current_umask();
req->in.h.opcode = FUSE_CREATE;
req->in.h.nodeid = get_node_id(dir);
req->in.numargs = 2;
req->in.args[0].size = fc->minor < 12 ? sizeof(struct fuse_open_in) :
sizeof(inarg);
req->in.args[0].value = &inarg;
req->in.args[1].size = entry->d_name.len + 1;
req->in.args[1].value = entry->d_name.name;
req->out.numargs = 2;
if (fc->minor < 9)
req->out.args[0].size = FUSE_COMPAT_ENTRY_OUT_SIZE;
else
req->out.args[0].size = sizeof(outentry);
req->out.args[0].value = &outentry;
req->out.args[1].size = sizeof(outopen);
req->out.args[1].value = &outopen;
fuse_request_send(fc, req);
err = req->out.h.error;
if (err) {
if (err == -ENOSYS)
fc->no_create = 1;
goto out_free_ff;
}
err = -EIO;
if (!S_ISREG(outentry.attr.mode) || invalid_nodeid(outentry.nodeid))
goto out_free_ff;
fuse_put_request(fc, req);
ff->fh = outopen.fh;
ff->nodeid = outentry.nodeid;
ff->open_flags = outopen.open_flags;
inode = fuse_iget(dir->i_sb, outentry.nodeid, outentry.generation,
&outentry.attr, entry_attr_timeout(&outentry), 0);
if (!inode) {
flags &= ~(O_CREAT | O_EXCL | O_TRUNC);
fuse_sync_release(ff, flags);
fuse_queue_forget(fc, forget, outentry.nodeid, 1);
return -ENOMEM;
}
kfree(forget);
d_instantiate(entry, inode);
fuse_change_entry_timeout(entry, &outentry);
fuse_invalidate_attr(dir);
file = lookup_instantiate_filp(nd, entry, generic_file_open);
if (IS_ERR(file)) {
fuse_sync_release(ff, flags);
return PTR_ERR(file);
}
file->private_data = fuse_file_get(ff);
fuse_finish_open(inode, file);
return 0;
out_free_ff:
fuse_file_free(ff);
out_put_request:
fuse_put_request(fc, req);
out_put_forget_req:
kfree(forget);
return err;
}
/*
* Atomic create+open operation
*
* If the filesystem doesn't support this, then fall back to separate
* 'mknod' + 'open' requests.
*/
static int fuse_create_open(struct inode *dir, struct dentry *entry, int mode,
struct nameidata *nd)
{
int err;
struct inode *inode;
struct fuse_conn *fc = get_fuse_conn(dir);
struct fuse_req *req;
struct fuse_req *forget_req;
struct fuse_open_in inarg;
struct fuse_open_out outopen;
struct fuse_entry_out outentry;
struct fuse_file *ff;
struct file *file;
int flags = nd->intent.open.flags - 1;
if (fc->no_create)
return -ENOSYS;
forget_req = fuse_get_req(fc);
if (IS_ERR(forget_req))
return PTR_ERR(forget_req);
req = fuse_get_req(fc);
err = PTR_ERR(req);
if (IS_ERR(req))
goto out_put_forget_req;
err = -ENOMEM;
ff = fuse_file_alloc();
if (!ff)
goto out_put_request;
flags &= ~O_NOCTTY;
memset(&inarg, 0, sizeof(inarg));
inarg.flags = flags;
inarg.mode = mode;
req->in.h.opcode = FUSE_CREATE;
req->in.h.nodeid = get_node_id(dir);
req->in.numargs = 2;
req->in.args[0].size = sizeof(inarg);
req->in.args[0].value = &inarg;
req->in.args[1].size = entry->d_name.len + 1;
req->in.args[1].value = entry->d_name.name;
req->out.numargs = 2;
req->out.args[0].size = sizeof(outentry);
req->out.args[0].value = &outentry;
req->out.args[1].size = sizeof(outopen);
req->out.args[1].value = &outopen;
request_send(fc, req);
err = req->out.h.error;
if (err) {
if (err == -ENOSYS)
fc->no_create = 1;
goto out_free_ff;
}
err = -EIO;
if (!S_ISREG(outentry.attr.mode) || invalid_nodeid(outentry.nodeid))
goto out_free_ff;
fuse_put_request(fc, req);
inode = fuse_iget(dir->i_sb, outentry.nodeid, outentry.generation,
&outentry.attr);
if (!inode) {
flags &= ~(O_CREAT | O_EXCL | O_TRUNC);
ff->fh = outopen.fh;
fuse_sync_release(fc, ff, outentry.nodeid, flags);
fuse_send_forget(fc, forget_req, outentry.nodeid, 1);
return -ENOMEM;
}
fuse_put_request(fc, forget_req);
d_instantiate(entry, inode);
fuse_invalidate_attr(dir);
fuse_change_timeout(entry, &outentry);
file = lookup_instantiate_filp(nd, entry, generic_file_open);
if (IS_ERR(file)) {
ff->fh = outopen.fh;
fuse_sync_release(fc, ff, outentry.nodeid, flags);
return PTR_ERR(file);
}
fuse_finish_open(inode, file, ff, &outopen);
return 0;
out_free_ff:
fuse_file_free(ff);
out_put_request:
fuse_put_request(fc, req);
out_put_forget_req:
fuse_put_request(fc, forget_req);
return err;
}
int ll_revalidate_it(struct dentry *de, int lookup_flags,
struct lookup_intent *it)
{
struct md_op_data *op_data;
struct ptlrpc_request *req = NULL;
struct lookup_intent lookup_it = { .it_op = IT_LOOKUP };
struct obd_export *exp;
struct inode *parent = de->d_parent->d_inode;
int rc;
ENTRY;
CDEBUG(D_VFSTRACE, "VFS Op:name=%s,intent=%s\n", de->d_name.name,
LL_IT2STR(it));
if (de->d_inode == NULL) {
__u64 ibits;
/* We can only use negative dentries if this is stat or lookup,
for opens and stuff we do need to query server. */
/* If there is IT_CREAT in intent op set, then we must throw
away this negative dentry and actually do the request to
kernel to create whatever needs to be created (if possible)*/
if (it && (it->it_op & IT_CREAT))
RETURN(0);
if (de->d_flags & DCACHE_LUSTRE_INVALID)
RETURN(0);
ibits = MDS_INODELOCK_UPDATE;
rc = ll_have_md_lock(parent, &ibits, LCK_MINMODE);
GOTO(out_sa, rc);
}
/* Never execute intents for mount points.
* Attributes will be fixed up in ll_inode_revalidate_it */
if (d_mountpoint(de))
GOTO(out_sa, rc = 1);
/* need to get attributes in case root got changed from other client */
if (de == de->d_sb->s_root) {
rc = __ll_inode_revalidate_it(de, it, MDS_INODELOCK_LOOKUP);
if (rc == 0)
rc = 1;
GOTO(out_sa, rc);
}
exp = ll_i2mdexp(de->d_inode);
OBD_FAIL_TIMEOUT(OBD_FAIL_MDC_REVALIDATE_PAUSE, 5);
ll_frob_intent(&it, &lookup_it);
LASSERT(it);
if (it->it_op == IT_LOOKUP && !(de->d_flags & DCACHE_LUSTRE_INVALID))
RETURN(1);
if ((it->it_op == IT_OPEN) && de->d_inode) {
struct inode *inode = de->d_inode;
struct ll_inode_info *lli = ll_i2info(inode);
struct obd_client_handle **och_p;
__u64 *och_usecount;
__u64 ibits;
/*
* We used to check for MDS_INODELOCK_OPEN here, but in fact
* just having LOOKUP lock is enough to justify inode is the
* same. And if inode is the same and we have suitable
* openhandle, then there is no point in doing another OPEN RPC
* just to throw away newly received openhandle. There are no
* security implications too, if file owner or access mode is
* change, LOOKUP lock is revoked.
*/
if (it->it_flags & FMODE_WRITE) {
och_p = &lli->lli_mds_write_och;
och_usecount = &lli->lli_open_fd_write_count;
} else if (it->it_flags & FMODE_EXEC) {
och_p = &lli->lli_mds_exec_och;
och_usecount = &lli->lli_open_fd_exec_count;
} else {
och_p = &lli->lli_mds_read_och;
och_usecount = &lli->lli_open_fd_read_count;
}
/* Check for the proper lock. */
ibits = MDS_INODELOCK_LOOKUP;
if (!ll_have_md_lock(inode, &ibits, LCK_MINMODE))
goto do_lock;
cfs_mutex_lock(&lli->lli_och_mutex);
if (*och_p) { /* Everything is open already, do nothing */
/*(*och_usecount)++; Do not let them steal our open
handle from under us */
SET_BUT_UNUSED(och_usecount);
/* XXX The code above was my original idea, but in case
we have the handle, but we cannot use it due to later
checks (e.g. O_CREAT|O_EXCL flags set), nobody
would decrement counter increased here. So we just
hope the lock won't be invalidated in between. But
if it would be, we'll reopen the open request to
MDS later during file open path */
cfs_mutex_unlock(&lli->lli_och_mutex);
RETURN(1);
} else {
cfs_mutex_unlock(&lli->lli_och_mutex);
}
}
if (it->it_op == IT_GETATTR) {
rc = ll_statahead_enter(parent, &de, 0);
if (rc == 1)
goto mark;
}
do_lock:
op_data = ll_prep_md_op_data(NULL, parent, de->d_inode,
de->d_name.name, de->d_name.len,
0, LUSTRE_OPC_ANY, NULL);
if (IS_ERR(op_data))
RETURN(PTR_ERR(op_data));
it->it_create_mode &= ~cfs_curproc_umask();
it->it_create_mode |= M_CHECK_STALE;
rc = md_intent_lock(exp, op_data, NULL, 0, it,
lookup_flags,
&req, ll_md_blocking_ast, 0);
it->it_create_mode &= ~M_CHECK_STALE;
ll_finish_md_op_data(op_data);
/* If req is NULL, then md_intent_lock only tried to do a lock match;
* if all was well, it will return 1 if it found locks, 0 otherwise. */
if (req == NULL && rc >= 0) {
if (!rc)
goto do_lookup;
GOTO(out, rc);
}
if (rc < 0) {
if (rc != -ESTALE) {
CDEBUG(D_INFO, "ll_intent_lock: rc %d : it->it_status "
"%d\n", rc, it->d.lustre.it_status);
}
GOTO(out, rc = 0);
}
revalidate_finish:
rc = ll_revalidate_it_finish(req, it, de);
if (rc != 0) {
if (rc != -ESTALE && rc != -ENOENT)
ll_intent_release(it);
GOTO(out, rc = 0);
}
if ((it->it_op & IT_OPEN) && de->d_inode &&
!S_ISREG(de->d_inode->i_mode) &&
!S_ISDIR(de->d_inode->i_mode)) {
ll_release_openhandle(de, it);
}
rc = 1;
/* unfortunately ll_intent_lock may cause a callback and revoke our
* dentry */
ll_dentry_rehash(de, 0);
out:
/* We do not free request as it may be reused during following lookup
* (see comment in mdc/mdc_locks.c::mdc_intent_lock()), request will
* be freed in ll_lookup_it or in ll_intent_release. But if
* request was not completed, we need to free it. (bug 5154, 9903) */
if (req != NULL && !it_disposition(it, DISP_ENQ_COMPLETE))
ptlrpc_req_finished(req);
if (rc == 0) {
ll_unhash_aliases(de->d_inode);
/* done in ll_unhash_aliases()
dentry->d_flags |= DCACHE_LUSTRE_INVALID; */
} else {
__u64 bits = 0;
CDEBUG(D_DENTRY, "revalidated dentry %.*s (%p) parent %p "
"inode %p refc %d\n", de->d_name.len,
de->d_name.name, de, de->d_parent, de->d_inode,
atomic_read(&de->d_count));
ll_set_lock_data(exp, de->d_inode, it, &bits);
ll_dentry_reset_flags(de, bits);
ll_lookup_finish_locks(it, de);
}
mark:
if (it != NULL && it->it_op == IT_GETATTR && rc > 0)
ll_statahead_mark(parent, de);
RETURN(rc);
/*
* This part is here to combat evil-evil race in real_lookup on 2.6
* kernels. The race details are: We enter do_lookup() looking for some
* name, there is nothing in dcache for this name yet and d_lookup()
* returns NULL. We proceed to real_lookup(), and while we do this,
* another process does open on the same file we looking up (most simple
* reproducer), open succeeds and the dentry is added. Now back to
* us. In real_lookup() we do d_lookup() again and suddenly find the
* dentry, so we call d_revalidate on it, but there is no lock, so
* without this code we would return 0, but unpatched real_lookup just
* returns -ENOENT in such a case instead of retrying the lookup. Once
* this is dealt with in real_lookup(), all of this ugly mess can go and
* we can just check locks in ->d_revalidate without doing any RPCs
* ever.
*/
do_lookup:
if (it != &lookup_it) {
/* MDS_INODELOCK_UPDATE needed for IT_GETATTR case. */
if (it->it_op == IT_GETATTR)
lookup_it.it_op = IT_GETATTR;
ll_lookup_finish_locks(it, de);
it = &lookup_it;
}
/* Do real lookup here. */
op_data = ll_prep_md_op_data(NULL, parent, NULL, de->d_name.name,
de->d_name.len, 0, (it->it_op & IT_CREAT ?
LUSTRE_OPC_CREATE :
LUSTRE_OPC_ANY), NULL);
if (IS_ERR(op_data))
RETURN(PTR_ERR(op_data));
rc = md_intent_lock(exp, op_data, NULL, 0, it, 0, &req,
ll_md_blocking_ast, 0);
if (rc >= 0) {
struct mdt_body *mdt_body;
struct lu_fid fid = {.f_seq = 0, .f_oid = 0, .f_ver = 0};
mdt_body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
if (de->d_inode)
fid = *ll_inode2fid(de->d_inode);
/* see if we got same inode, if not - return error */
if (lu_fid_eq(&fid, &mdt_body->fid1)) {
ll_finish_md_op_data(op_data);
op_data = NULL;
goto revalidate_finish;
}
ll_intent_release(it);
}
ll_finish_md_op_data(op_data);
GOTO(out, rc = 0);
out_sa:
/*
* For rc == 1 case, should not return directly to prevent losing
* statahead windows; for rc == 0 case, the "lookup" will be done later.
*/
if (it != NULL && it->it_op == IT_GETATTR && rc == 1)
ll_statahead_enter(parent, &de, 1);
goto mark;
}
int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd)
{
int rc;
ENTRY;
if (nd && !(nd->flags & (LOOKUP_CONTINUE|LOOKUP_PARENT))) {
struct lookup_intent *it;
it = ll_convert_intent(&nd->intent.open, nd->flags);
if (IS_ERR(it))
RETURN(0);
if (it->it_op == (IT_OPEN|IT_CREAT) &&
nd->intent.open.flags & O_EXCL) {
CDEBUG(D_VFSTRACE, "create O_EXCL, returning 0\n");
rc = 0;
goto out_it;
}
rc = ll_revalidate_it(dentry, nd->flags, it);
if (rc && (nd->flags & LOOKUP_OPEN) &&
it_disposition(it, DISP_OPEN_OPEN)) {/*Open*/
// XXX Code duplication with ll_lookup_nd
if (S_ISFIFO(dentry->d_inode->i_mode)) {
// We cannot call open here as it would
// deadlock.
ptlrpc_req_finished(
(struct ptlrpc_request *)
it->d.lustre.it_data);
} else {
#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,17))
/* 2.6.1[456] have a bug in open_namei() that forgets to check
* nd->intent.open.file for error, so we need to return it as lookup's result
* instead */
struct file *filp;
nd->intent.open.file->private_data = it;
filp = lookup_instantiate_filp(nd, dentry,NULL);
if (IS_ERR(filp)) {
rc = PTR_ERR(filp);
}
#else
nd->intent.open.file->private_data = it;
(void)lookup_instantiate_filp(nd, dentry,NULL);
#endif
}
}
if (!rc && (nd->flags & LOOKUP_CREATE) &&
it_disposition(it, DISP_OPEN_CREATE)) {
/* We created something but we may only return
* negative dentry here, so save request in dentry,
* if lookup will be called later on, it will
* pick the request, otherwise it would be freed
* with dentry */
ll_d2d(dentry)->lld_it = it;
it = NULL; /* avoid freeing */
}
out_it:
if (it) {
ll_intent_release(it);
OBD_FREE(it, sizeof(*it));
}
} else {
rc = ll_revalidate_it(dentry, 0, NULL);
}
RETURN(rc);
}
