[PATCH] v9fs: fix atomic create open
Latchesar Ionkov [Thu, 2 Mar 2006 10:54:30 +0000 (02:54 -0800)]
In order to assure atomic create+open v9fs stores the open fid produced by
v9fs_vfs_create in the dentry, from where v9fs_file_open retrieves it and
associates it with the open file.

This patch modifies v9fs to use nameidata.intent.open values to do the atomic
create+open.

Signed-off-by: Latchesar Ionkov <lucho@ionkov.net>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

fs/9p/fid.c
fs/9p/fid.h
fs/9p/v9fs_vfs.h
fs/9p/vfs_file.c
fs/9p/vfs_inode.c
fs/9p/vfs_super.c

index eda4497..c27f546 100644 (file)
@@ -40,7 +40,7 @@
  *
  */
 
-static int v9fs_fid_insert(struct v9fs_fid *fid, struct dentry *dentry)
+int v9fs_fid_insert(struct v9fs_fid *fid, struct dentry *dentry)
 {
        struct list_head *fid_list = (struct list_head *)dentry->d_fsdata;
        dprintk(DEBUG_9P, "fid %d (%p) dentry %s (%p)\n", fid->fid, fid,
@@ -68,14 +68,11 @@ static int v9fs_fid_insert(struct v9fs_fid *fid, struct dentry *dentry)
  *
  */
 
-struct v9fs_fid *v9fs_fid_create(struct dentry *dentry,
-       struct v9fs_session_info *v9ses, int fid, int create)
+struct v9fs_fid *v9fs_fid_create(struct v9fs_session_info *v9ses, int fid)
 {
        struct v9fs_fid *new;
 
-       dprintk(DEBUG_9P, "fid create dentry %p, fid %d, create %d\n",
-               dentry, fid, create);
-
+       dprintk(DEBUG_9P, "fid create fid %d\n", fid);
        new = kmalloc(sizeof(struct v9fs_fid), GFP_KERNEL);
        if (new == NULL) {
                dprintk(DEBUG_ERROR, "Out of Memory\n");
@@ -85,19 +82,13 @@ struct v9fs_fid *v9fs_fid_create(struct dentry *dentry,
        new->fid = fid;
        new->v9ses = v9ses;
        new->fidopen = 0;
-       new->fidcreate = create;
        new->fidclunked = 0;
        new->iounit = 0;
        new->rdir_pos = 0;
        new->rdir_fcall = NULL;
+       INIT_LIST_HEAD(&new->list);
 
-       if (v9fs_fid_insert(new, dentry) == 0)
                return new;
-       else {
-               dprintk(DEBUG_ERROR, "Problems inserting to dentry\n");
-               kfree(new);
-               return NULL;
-       }
 }
 
 /**
@@ -119,7 +110,7 @@ void v9fs_fid_destroy(struct v9fs_fid *fid)
 static struct v9fs_fid *v9fs_fid_walk_up(struct dentry *dentry)
 {
        int fidnum, cfidnum, err;
-       struct v9fs_fid *cfid;
+       struct v9fs_fid *cfid, *fid;
        struct dentry *cde;
        struct v9fs_session_info *v9ses;
 
@@ -158,7 +149,16 @@ static struct v9fs_fid *v9fs_fid_walk_up(struct dentry *dentry)
                cde = cde->d_parent;
        }
 
-       return v9fs_fid_create(dentry, v9ses, fidnum, 0);
+       fid = v9fs_fid_create(v9ses, fidnum);
+       if (fid) {
+               err = v9fs_fid_insert(fid, dentry);
+               if (err < 0) {
+                       kfree(fid);
+                       goto clunk_fid;
+               }
+       }
+
+       return fid;
 
 clunk_fid:
        v9fs_t_clunk(v9ses, fidnum);
@@ -179,29 +179,12 @@ clunk_fid:
 struct v9fs_fid *v9fs_fid_lookup(struct dentry *dentry)
 {
        struct list_head *fid_list = (struct list_head *)dentry->d_fsdata;
-       struct v9fs_fid *current_fid = NULL;
-       struct v9fs_fid *temp = NULL;
        struct v9fs_fid *return_fid = NULL;
 
        dprintk(DEBUG_9P, " dentry: %s (%p)\n", dentry->d_iname, dentry);
 
-       if (fid_list) {
-               list_for_each_entry_safe(current_fid, temp, fid_list, list) {
-                       if (!current_fid->fidcreate) {
-                               return_fid = current_fid;
-                               break;
-                       }
-               }
-
-               if (!return_fid)
-                       return_fid = current_fid;
-       }
-
-       /* we are at the root but didn't match */
-       if ((!return_fid) && (dentry->d_parent == dentry)) {
-               /* TODO: clone attach with new uid */
-               return_fid = current_fid;
-       }
+       if (fid_list)
+               return_fid = list_entry(fid_list->next, struct v9fs_fid, list);
 
        if (!return_fid) {
                struct dentry *par = current->fs->pwd->d_parent;
@@ -228,25 +211,3 @@ struct v9fs_fid *v9fs_fid_lookup(struct dentry *dentry)
 
        return return_fid;
 }
-
-struct v9fs_fid *v9fs_fid_get_created(struct dentry *dentry)
-{
-       struct list_head *fid_list;
-       struct v9fs_fid *fid, *ftmp, *ret;
-
-       dprintk(DEBUG_9P, " dentry: %s (%p)\n", dentry->d_iname, dentry);
-       fid_list = (struct list_head *)dentry->d_fsdata;
-       ret = NULL;
-       if (fid_list) {
-               list_for_each_entry_safe(fid, ftmp, fid_list, list) {
-                       if (fid->fidcreate && fid->pid == current->pid) {
-                               list_del(&fid->list);
-                               ret = fid;
-                               break;
-                       }
-               }
-       }
-
-       dprintk(DEBUG_9P, "return %p\n", ret);
-       return ret;
-}
index 84c673a..7ccf0d0 100644 (file)
@@ -33,7 +33,6 @@ struct v9fs_fid {
 
        u32 fid;
        unsigned char fidopen;    /* set when fid is opened */
-       unsigned char fidcreate;  /* set when fid was just created */
        unsigned char fidclunked; /* set when fid has already been clunked */
 
        struct v9fs_qid qid;
@@ -56,5 +55,5 @@ struct v9fs_fid {
 struct v9fs_fid *v9fs_fid_lookup(struct dentry *dentry);
 struct v9fs_fid *v9fs_fid_get_created(struct dentry *);
 void v9fs_fid_destroy(struct v9fs_fid *fid);
-struct v9fs_fid *v9fs_fid_create(struct dentry *,
-       struct v9fs_session_info *v9ses, int fid, int create);
+struct v9fs_fid *v9fs_fid_create(struct v9fs_session_info *, int fid);
+int v9fs_fid_insert(struct v9fs_fid *fid, struct dentry *dentry);
index 69cf290..a759278 100644 (file)
@@ -51,3 +51,4 @@ int v9fs_dir_release(struct inode *inode, struct file *filp);
 int v9fs_file_open(struct inode *inode, struct file *file);
 void v9fs_inode2stat(struct inode *inode, struct v9fs_stat *stat);
 void v9fs_dentry_release(struct dentry *);
+int v9fs_uflags2omode(int uflags);
index c7e14d9..de3a129 100644 (file)
 int v9fs_file_open(struct inode *inode, struct file *file)
 {
        struct v9fs_session_info *v9ses = v9fs_inode2v9ses(inode);
-       struct v9fs_fid *v9fid, *fid;
+       struct v9fs_fid *vfid;
        struct v9fs_fcall *fcall = NULL;
-       int open_mode = 0;
-       unsigned int iounit = 0;
-       int newfid = -1;
-       long result = -1;
+       int omode;
+       int fid = V9FS_NOFID;
+       int err;
 
        dprintk(DEBUG_VFS, "inode: %p file: %p \n", inode, file);
 
-       v9fid = v9fs_fid_get_created(file->f_dentry);
-       if (!v9fid)
-               v9fid = v9fs_fid_lookup(file->f_dentry);
-
-       if (!v9fid) {
+       vfid = v9fs_fid_lookup(file->f_dentry);
+       if (!vfid) {
                dprintk(DEBUG_ERROR, "Couldn't resolve fid from dentry\n");
                return -EBADF;
        }
 
-       if (!v9fid->fidcreate) {
-               fid = kmalloc(sizeof(struct v9fs_fid), GFP_KERNEL);
-               if (fid == NULL) {
-                       dprintk(DEBUG_ERROR, "Out of Memory\n");
-                       return -ENOMEM;
-               }
-
-               fid->fidopen = 0;
-               fid->fidcreate = 0;
-               fid->fidclunked = 0;
-               fid->iounit = 0;
-               fid->v9ses = v9ses;
-
-               newfid = v9fs_get_idpool(&v9ses->fidpool);
-               if (newfid < 0) {
+       fid = v9fs_get_idpool(&v9ses->fidpool);
+       if (fid < 0) {
                        eprintk(KERN_WARNING, "newfid fails!\n");
                        return -ENOSPC;
                }
 
-               result =
-                   v9fs_t_walk(v9ses, v9fid->fid, newfid, NULL, NULL);
-
-               if (result < 0) {
-                       v9fs_put_idpool(newfid, &v9ses->fidpool);
+       err = v9fs_t_walk(v9ses, vfid->fid, fid, NULL, NULL);
+       if (err < 0) {
                        dprintk(DEBUG_ERROR, "rewalk didn't work\n");
-                       return -EBADF;
+               goto put_fid;
+       }
+
+       vfid = kmalloc(sizeof(struct v9fs_fid), GFP_KERNEL);
+       if (vfid == NULL) {
+               dprintk(DEBUG_ERROR, "out of memory\n");
+               goto clunk_fid;
                }
 
-               fid->fid = newfid;
-               v9fid = fid;
                /* TODO: do special things for O_EXCL, O_NOFOLLOW, O_SYNC */
                /* translate open mode appropriately */
-               open_mode = file->f_flags & 0x3;
+       omode = v9fs_uflags2omode(file->f_flags);
+       err = v9fs_t_open(v9ses, fid, omode, &fcall);
+       if (err < 0) {
+               PRINT_FCALL_ERROR("open failed", fcall);
+               goto destroy_vfid;
+       }
 
-               if (file->f_flags & O_EXCL)
-                       open_mode |= V9FS_OEXCL;
+       file->private_data = vfid;
+       vfid->fid = fid;
+       vfid->fidopen = 1;
+       vfid->fidclunked = 0;
+       vfid->iounit = fcall->params.ropen.iounit;
+       vfid->rdir_pos = 0;
+       vfid->rdir_fcall = NULL;
+       vfid->filp = file;
+       kfree(fcall);
 
-               if (v9ses->extended) {
-                       if (file->f_flags & O_TRUNC)
-                               open_mode |= V9FS_OTRUNC;
+       return 0;
 
-                       if (file->f_flags & O_APPEND)
-                               open_mode |= V9FS_OAPPEND;
-               }
+destroy_vfid:
+       v9fs_fid_destroy(vfid);
 
-               result = v9fs_t_open(v9ses, newfid, open_mode, &fcall);
-               if (result < 0) {
-                       PRINT_FCALL_ERROR("open failed", fcall);
-                       kfree(fcall);
-                       return result;
-               }
+clunk_fid:
+       v9fs_t_clunk(v9ses, fid);
 
-               iounit = fcall->params.ropen.iounit;
+put_fid:
+       v9fs_put_idpool(fid, &v9ses->fidpool);
                kfree(fcall);
-       } else {
-               /* create case */
-               newfid = v9fid->fid;
-               iounit = v9fid->iounit;
-               v9fid->fidcreate = 0;
-       }
-
-       file->private_data = v9fid;
-
-       v9fid->rdir_pos = 0;
-       v9fid->rdir_fcall = NULL;
-       v9fid->fidopen = 1;
-       v9fid->filp = file;
-       v9fid->iounit = iounit;
 
-       return 0;
+       return err;
 }
 
 /**
@@ -289,9 +265,7 @@ v9fs_file_write(struct file *filp, const char __user * data,
                total += result;
        } while (count);
 
-       if(inode->i_mapping->nrpages)
                invalidate_inode_pages2(inode->i_mapping);
-
        return total;
 }
 
index 63e5b03..dce729d 100644 (file)
@@ -125,6 +125,38 @@ static int p9mode2unixmode(struct v9fs_session_info *v9ses, int mode)
        return res;
 }
 
+int v9fs_uflags2omode(int uflags)
+{
+       int ret;
+
+       ret = 0;
+       switch (uflags&3) {
+       default:
+       case O_RDONLY:
+               ret = V9FS_OREAD;
+               break;
+
+       case O_WRONLY:
+               ret = V9FS_OWRITE;
+               break;
+
+       case O_RDWR:
+               ret = V9FS_ORDWR;
+               break;
+       }
+
+       if (uflags & O_EXCL)
+               ret |= V9FS_OEXCL;
+
+       if (uflags & O_TRUNC)
+               ret |= V9FS_OTRUNC;
+
+       if (uflags & O_APPEND)
+               ret |= V9FS_OAPPEND;
+
+       return ret;
+}
+
 /**
  * v9fs_blank_wstat - helper function to setup a 9P stat structure
  * @v9ses: 9P session info (for determining extended mode)
@@ -163,7 +195,7 @@ v9fs_blank_wstat(struct v9fs_wstat *wstat)
 
 struct inode *v9fs_get_inode(struct super_block *sb, int mode)
 {
-       struct inode *inode = NULL;
+       struct inode *inode;
        struct v9fs_session_info *v9ses = sb->s_fs_info;
 
        dprintk(DEBUG_VFS, "super block: %p mode: %o\n", sb, mode);
@@ -222,171 +254,135 @@ struct inode *v9fs_get_inode(struct super_block *sb, int mode)
        return inode;
 }
 
-/**
- * v9fs_create - helper function to create files and directories
- * @dir: directory inode file is being created in
- * @file_dentry: dentry file is being created in
- * @perm: permissions file is being created with
- * @open_mode: resulting open mode for file
- *
- */
-
 static int
-v9fs_create(struct inode *dir,
-           struct dentry *file_dentry,
-           unsigned int perm, unsigned int open_mode)
+v9fs_create(struct v9fs_session_info *v9ses, u32 pfid, char *name,
+       u32 perm, u8 mode, u32 *fidp, struct v9fs_qid *qid, u32 *iounit)
 {
-       struct v9fs_session_info *v9ses = v9fs_inode2v9ses(dir);
-       struct super_block *sb = dir->i_sb;
-       struct v9fs_fid *dirfid =
-           v9fs_fid_lookup(file_dentry->d_parent);
-       struct v9fs_fid *fid = NULL;
-       struct inode *file_inode = NULL;
-       struct v9fs_fcall *fcall = NULL;
-       struct v9fs_qid qid;
-       int dirfidnum = -1;
-       long newfid = -1;
-       int result = 0;
-       unsigned int iounit = 0;
-       int wfidno = -1;
+       u32 fid;
        int err;
+       struct v9fs_fcall *fcall;
 
-       perm = unixmode2p9mode(v9ses, perm);
-
-       dprintk(DEBUG_VFS, "dir: %p dentry: %p perm: %o mode: %o\n", dir,
-               file_dentry, perm, open_mode);
-
-       if (!dirfid)
-               return -EBADF;
-
-       dirfidnum = dirfid->fid;
-       if (dirfidnum < 0) {
-               dprintk(DEBUG_ERROR, "No fid for the directory #%lu\n",
-                       dir->i_ino);
-               return -EBADF;
-       }
-
-       if (file_dentry->d_inode) {
-               dprintk(DEBUG_ERROR,
-                       "Odd. There is an inode for dir %lu, name :%s:\n",
-                       dir->i_ino, file_dentry->d_name.name);
-               return -EEXIST;
-       }
-
-       newfid = v9fs_get_idpool(&v9ses->fidpool);
-       if (newfid < 0) {
+       fid = v9fs_get_idpool(&v9ses->fidpool);
+       if (fid < 0) {
                eprintk(KERN_WARNING, "no free fids available\n");
-               return -ENOSPC;
+               err = -ENOSPC;
+               goto error;
        }
 
-       result = v9fs_t_walk(v9ses, dirfidnum, newfid, NULL, &fcall);
-       if (result < 0) {
+       err = v9fs_t_walk(v9ses, pfid, fid, NULL, &fcall);
+       if (err < 0) {
                PRINT_FCALL_ERROR("clone error", fcall);
-               v9fs_put_idpool(newfid, &v9ses->fidpool);
-               newfid = -1;
-               goto CleanUpFid;
+               goto error;
        }
-
        kfree(fcall);
-       fcall = NULL;
 
-       result = v9fs_t_create(v9ses, newfid, (char *)file_dentry->d_name.name,
-                              perm, open_mode, &fcall);
-       if (result < 0) {
+       err = v9fs_t_create(v9ses, fid, name, perm, mode, &fcall);
+       if (err < 0) {
                PRINT_FCALL_ERROR("create fails", fcall);
-               goto CleanUpFid;
+               goto error;
        }
 
-       iounit = fcall->params.rcreate.iounit;
-       qid = fcall->params.rcreate.qid;
+       if (iounit)
+               *iounit = fcall->params.rcreate.iounit;
+
+       if (qid)
+               *qid = fcall->params.rcreate.qid;
+
+       if (fidp)
+               *fidp = fid;
+
        kfree(fcall);
-       fcall = NULL;
+       return 0;
 
-       if (!(perm&V9FS_DMDIR)) {
-               fid = v9fs_fid_create(file_dentry, v9ses, newfid, 1);
-               dprintk(DEBUG_VFS, "fid %p %d\n", fid, fid->fidcreate);
-               if (!fid) {
-                       result = -ENOMEM;
-                       goto CleanUpFid;
-               }
+error:
+       if (fid >= 0)
+               v9fs_put_idpool(fid, &v9ses->fidpool);
 
-               fid->qid = qid;
-               fid->iounit = iounit;
-       } else {
-               err = v9fs_t_clunk(v9ses, newfid);
-               newfid = -1;
-               if (err < 0)
-                       dprintk(DEBUG_ERROR, "clunk for mkdir failed: %d\n", err);
-       }
+       kfree(fcall);
+       return err;
+}
+
+static struct v9fs_fid*
+v9fs_clone_walk(struct v9fs_session_info *v9ses, u32 fid, struct dentry *dentry)
+{
+       int err;
+       u32 nfid;
+       struct v9fs_fid *ret;
+       struct v9fs_fcall *fcall;
 
-       /* walk to the newly created file and put the fid in the dentry */
-       wfidno = v9fs_get_idpool(&v9ses->fidpool);
-       if (wfidno < 0) {
+       nfid = v9fs_get_idpool(&v9ses->fidpool);
+       if (nfid < 0) {
                eprintk(KERN_WARNING, "no free fids available\n");
-               return -ENOSPC;
+               err = -ENOSPC;
+               goto error;
        }
 
-       result = v9fs_t_walk(v9ses, dirfidnum, wfidno,
-               (char *) file_dentry->d_name.name, &fcall);
-       if (result < 0) {
-               PRINT_FCALL_ERROR("clone error", fcall);
-               v9fs_put_idpool(wfidno, &v9ses->fidpool);
-               wfidno = -1;
-               goto CleanUpFid;
+       err = v9fs_t_walk(v9ses, fid, nfid, (char *) dentry->d_name.name,
+               &fcall);
+
+       if (err < 0) {
+               PRINT_FCALL_ERROR("walk error", fcall);
+               v9fs_put_idpool(nfid, &v9ses->fidpool);
+               goto error;
        }
+
        kfree(fcall);
        fcall = NULL;
+       ret = v9fs_fid_create(v9ses, nfid);
+       if (!ret) {
+               err = -ENOMEM;
+               goto clunk_fid;
+       }
 
-       if (!v9fs_fid_create(file_dentry, v9ses, wfidno, 0)) {
-               v9fs_put_idpool(wfidno, &v9ses->fidpool);
-
-               goto CleanUpFid;
+       err = v9fs_fid_insert(ret, dentry);
+       if (err < 0) {
+               v9fs_fid_destroy(ret);
+               goto clunk_fid;
        }
 
-       if ((perm & V9FS_DMSYMLINK) || (perm & V9FS_DMLINK) ||
-           (perm & V9FS_DMNAMEDPIPE) || (perm & V9FS_DMSOCKET) ||
-           (perm & V9FS_DMDEVICE))
-               return 0;
+       return ret;
 
-       result = v9fs_t_stat(v9ses, wfidno, &fcall);
-       if (result < 0) {
-               PRINT_FCALL_ERROR("stat error", fcall);
-               goto CleanUpFid;
-       }
+clunk_fid:
+       v9fs_t_clunk(v9ses, nfid);
+
+error:
+       kfree(fcall);
+       return ERR_PTR(err);
+}
 
+struct inode *
+v9fs_inode_from_fid(struct v9fs_session_info *v9ses, u32 fid,
+       struct super_block *sb)
+{
+       int err, umode;
+       struct inode *ret;
+       struct v9fs_fcall *fcall;
 
-       file_inode = v9fs_get_inode(sb,
-               p9mode2unixmode(v9ses, fcall->params.rstat.stat.mode));
+       ret = NULL;
+       err = v9fs_t_stat(v9ses, fid, &fcall);
+       if (err) {
+               PRINT_FCALL_ERROR("stat error", fcall);
+               goto error;
+       }
 
-       if ((!file_inode) || IS_ERR(file_inode)) {
-               dprintk(DEBUG_ERROR, "create inode failed\n");
-               result = -EBADF;
-               goto CleanUpFid;
+       umode = p9mode2unixmode(v9ses, fcall->params.rstat.stat.mode);
+       ret = v9fs_get_inode(sb, umode);
+       if (IS_ERR(ret)) {
+               err = PTR_ERR(ret);
+               ret = NULL;
+               goto error;
        }
 
-       v9fs_stat2inode(&fcall->params.rstat.stat, file_inode, sb);
+       v9fs_stat2inode(&fcall->params.rstat.stat, ret, sb);
        kfree(fcall);
-       fcall = NULL;
-       file_dentry->d_op = &v9fs_dentry_operations;
-       d_instantiate(file_dentry, file_inode);
-
-       return 0;
+       return ret;
 
-      CleanUpFid:
+error:
        kfree(fcall);
-       fcall = NULL;
+       if (ret)
+               iput(ret);
 
-       if (newfid >= 0) {
-               err = v9fs_t_clunk(v9ses, newfid);
-               if (err < 0)
-                       dprintk(DEBUG_ERROR, "clunk failed: %d\n", err);
-       }
-       if (wfidno >= 0) {
-               err = v9fs_t_clunk(v9ses, wfidno);
-               if (err < 0)
-                       dprintk(DEBUG_ERROR, "clunk failed: %d\n", err);
-       }
-       return result;
+       return ERR_PTR(err);
 }
 
 /**
@@ -440,20 +436,97 @@ static int v9fs_remove(struct inode *dir, struct dentry *file, int rmdir)
        return result;
 }
 
+static int
+v9fs_open_created(struct inode *inode, struct file *file)
+{
+       return 0;
+}
+
 /**
  * v9fs_vfs_create - VFS hook to create files
  * @inode: directory inode that is being deleted
  * @dentry:  dentry that is being deleted
- * @perm: create permissions
+ * @mode: create permissions
  * @nd: path information
  *
  */
 
 static int
-v9fs_vfs_create(struct inode *inode, struct dentry *dentry, int perm,
+v9fs_vfs_create(struct inode *dir, struct dentry *dentry, int mode,
                struct nameidata *nd)
 {
-       return v9fs_create(inode, dentry, perm, O_RDWR);
+       int err;
+       u32 fid, perm, iounit;
+       int flags;
+       struct v9fs_session_info *v9ses;
+       struct v9fs_fid *dfid, *vfid, *ffid;
+       struct inode *inode;
+       struct v9fs_qid qid;
+       struct file *filp;
+
+       inode = NULL;
+       vfid = NULL;
+       v9ses = v9fs_inode2v9ses(dir);
+       dfid = v9fs_fid_lookup(dentry->d_parent);
+       perm = unixmode2p9mode(v9ses, mode);
+
+       if (nd && nd->flags & LOOKUP_OPEN)
+               flags = nd->intent.open.flags - 1;
+       else
+               flags = O_RDWR;
+
+       err = v9fs_create(v9ses, dfid->fid, (char *) dentry->d_name.name,
+               perm, v9fs_uflags2omode(flags), &fid, &qid, &iounit);
+
+       if (err)
+               goto error;
+
+       vfid = v9fs_clone_walk(v9ses, dfid->fid, dentry);
+       if (IS_ERR(vfid)) {
+               err = PTR_ERR(vfid);
+               vfid = NULL;
+               goto error;
+       }
+
+       inode = v9fs_inode_from_fid(v9ses, vfid->fid, dir->i_sb);
+       if (IS_ERR(inode)) {
+               err = PTR_ERR(inode);
+               inode = NULL;
+               goto error;
+       }
+
+       dentry->d_op = &v9fs_dentry_operations;
+       d_instantiate(dentry, inode);
+
+       if (nd && nd->flags & LOOKUP_OPEN) {
+               ffid = v9fs_fid_create(v9ses, fid);
+               if (!ffid)
+                       return -ENOMEM;
+
+               filp = lookup_instantiate_filp(nd, dentry, v9fs_open_created);
+               if (IS_ERR(filp)) {
+                       v9fs_fid_destroy(ffid);
+                       return PTR_ERR(filp);
+               }
+
+               ffid->rdir_pos = 0;
+               ffid->rdir_fcall = NULL;
+               ffid->fidopen = 1;
+               ffid->iounit = iounit;
+               ffid->filp = filp;
+               filp->private_data = ffid;
+       }
+
+       return 0;
+
+error:
+       if (vfid)
+               v9fs_fid_destroy(vfid);
+
+       if (inode)
+               iput(inode);
+
+       return err;
 }
 
 /**
@@ -464,9 +537,57 @@ v9fs_vfs_create(struct inode *inode, struct dentry *dentry, int perm,
  *
  */
 
-static int v9fs_vfs_mkdir(struct inode *inode, struct dentry *dentry, int mode)
+static int v9fs_vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 {
-       return v9fs_create(inode, dentry, mode | S_IFDIR, O_RDONLY);
+       int err;
+       u32 fid, perm;
+       struct v9fs_session_info *v9ses;
+       struct v9fs_fid *dfid, *vfid;
+       struct inode *inode;
+
+       inode = NULL;
+       vfid = NULL;
+       v9ses = v9fs_inode2v9ses(dir);
+       dfid = v9fs_fid_lookup(dentry->d_parent);
+       perm = unixmode2p9mode(v9ses, mode | S_IFDIR);
+
+       err = v9fs_create(v9ses, dfid->fid, (char *) dentry->d_name.name,
+               perm, V9FS_OREAD, &fid, NULL, NULL);
+
+       if (err) {
+               dprintk(DEBUG_ERROR, "create error %d\n", err);
+               goto error;
+       }
+
+       err = v9fs_t_clunk(v9ses, fid);
+       if (err) {
+               dprintk(DEBUG_ERROR, "clunk error %d\n", err);
+               goto error;
+       }
+
+       vfid = v9fs_clone_walk(v9ses, dfid->fid, dentry);
+       if (IS_ERR(vfid)) {
+               err = PTR_ERR(vfid);
+               vfid = NULL;
+               goto error;
+       }
+
+       inode = v9fs_inode_from_fid(v9ses, vfid->fid, dir->i_sb);
+       if (IS_ERR(inode)) {
+               err = PTR_ERR(inode);
+               inode = NULL;
+               goto error;
+       }
+
+       dentry->d_op = &v9fs_dentry_operations;
+       d_instantiate(dentry, inode);
+       return 0;
+
+error:
+       if (vfid)
+               v9fs_fid_destroy(vfid);
+
+       return err;
 }
 
 /**
@@ -516,9 +637,8 @@ static struct dentry *v9fs_vfs_lookup(struct inode *dir, struct dentry *dentry,
                return ERR_PTR(-ENOSPC);
        }
 
-       result =
-           v9fs_t_walk(v9ses, dirfidnum, newfid, (char *)dentry->d_name.name,
-                       NULL);
+       result = v9fs_t_walk(v9ses, dirfidnum, newfid,
+               (char *)dentry->d_name.name, NULL);
        if (result < 0) {
                v9fs_put_idpool(newfid, &v9ses->fidpool);
                if (result == -ENOENT) {
@@ -551,13 +671,17 @@ static struct dentry *v9fs_vfs_lookup(struct inode *dir, struct dentry *dentry,
 
        inode->i_ino = v9fs_qid2ino(&fcall->params.rstat.stat.qid);
 
-       fid = v9fs_fid_create(dentry, v9ses, newfid, 0);
+       fid = v9fs_fid_create(v9ses, newfid);
        if (fid == NULL) {
                dprintk(DEBUG_ERROR, "couldn't insert\n");
                result = -ENOMEM;
                goto FreeFcall;
        }
 
+       result = v9fs_fid_insert(fid, dentry);
+       if (result < 0)
+               goto FreeFcall;
+
        fid->qid = fcall->params.rstat.stat.qid;
 
        dentry->d_op = &v9fs_dentry_operations;
@@ -886,8 +1010,8 @@ static int v9fs_readlink(struct dentry *dentry, char *buffer, int buflen)
        }
 
        /* copy extension buffer into buffer */
-       if (fcall->params.rstat.stat.extension.len+1 < buflen)
-               buflen = fcall->params.rstat.stat.extension.len + 1;
+       if (fcall->params.rstat.stat.extension.len < buflen)
+               buflen = fcall->params.rstat.stat.extension.len;
 
        memcpy(buffer, fcall->params.rstat.stat.extension.str, buflen - 1);
        buffer[buflen-1] = 0;
@@ -951,7 +1075,7 @@ static void *v9fs_vfs_follow_link(struct dentry *dentry, struct nameidata *nd)
        if (!link)
                link = ERR_PTR(-ENOMEM);
        else {
-               len = v9fs_readlink(dentry, link, PATH_MAX);
+               len = v9fs_readlink(dentry, link, strlen(link));
 
                if (len < 0) {
                        __putname(link);
@@ -983,53 +1107,75 @@ static void v9fs_vfs_put_link(struct dentry *dentry, struct nameidata *nd, void
 static int v9fs_vfs_mkspecial(struct inode *dir, struct dentry *dentry,
        int mode, const char *extension)
 {
-       int err, retval;
+       int err;
+       u32 fid, perm;
        struct v9fs_session_info *v9ses;
+       struct v9fs_fid *dfid, *vfid;
+       struct inode *inode;
        struct v9fs_fcall *fcall;
-       struct v9fs_fid *fid;
        struct v9fs_wstat wstat;
 
-       v9ses = v9fs_inode2v9ses(dir);
-       retval = -EPERM;
        fcall = NULL;
+       inode = NULL;
+       vfid = NULL;
+       v9ses = v9fs_inode2v9ses(dir);
+       dfid = v9fs_fid_lookup(dentry->d_parent);
+       perm = unixmode2p9mode(v9ses, mode);
 
        if (!v9ses->extended) {
                dprintk(DEBUG_ERROR, "not extended\n");
-               goto free_mem;
+               return -EPERM;
        }
 
-       /* issue a create */
-       retval = v9fs_create(dir, dentry, mode, 0);
-       if (retval != 0)
-               goto free_mem;
+       err = v9fs_create(v9ses, dfid->fid, (char *) dentry->d_name.name,
+               perm, V9FS_OREAD, &fid, NULL, NULL);
 
-       fid = v9fs_fid_get_created(dentry);
-       if (!fid) {
-               dprintk(DEBUG_ERROR, "couldn't resolve fid from dentry\n");
-               goto free_mem;
+       if (err)
+               goto error;
+
+       err = v9fs_t_clunk(v9ses, fid);
+       if (err)
+               goto error;
+
+       vfid = v9fs_clone_walk(v9ses, dfid->fid, dentry);
+       if (IS_ERR(vfid)) {
+               err = PTR_ERR(vfid);
+               vfid = NULL;
+               goto error;
+       }
+
+       inode = v9fs_inode_from_fid(v9ses, vfid->fid, dir->i_sb);
+       if (IS_ERR(inode)) {
+               err = PTR_ERR(inode);
+               inode = NULL;
+               goto error;
        }
 
        /* issue a Twstat */
        v9fs_blank_wstat(&wstat);
        wstat.muid = v9ses->name;
        wstat.extension = (char *) extension;
-       retval = v9fs_t_wstat(v9ses, fid->fid, &wstat, &fcall);
-       if (retval < 0) {
-               PRINT_FCALL_ERROR("wstat error", fcall);
-               goto free_mem;
-       }
-
-       err = v9fs_t_clunk(v9ses, fid->fid);
+       err = v9fs_t_wstat(v9ses, vfid->fid, &wstat, &fcall);
        if (err < 0) {
-               dprintk(DEBUG_ERROR, "clunk failed: %d\n", err);
-               goto free_mem;
+               PRINT_FCALL_ERROR("wstat error", fcall);
+               goto error;
        }
 
-       d_drop(dentry);         /* FID - will this also clunk? */
+       kfree(fcall);
+       dentry->d_op = &v9fs_dentry_operations;
+       d_instantiate(dentry, inode);
+       return 0;
 
-free_mem:
+error:
        kfree(fcall);
-       return retval;
+       if (vfid)
+               v9fs_fid_destroy(vfid);
+
+       if (inode)
+               iput(inode);
+
+       return err;
+
 }
 
 /**
index 2c4fa75..0c85872 100644 (file)
@@ -146,7 +146,6 @@ static struct super_block *v9fs_get_sb(struct file_system_type
        inode->i_gid = gid;
 
        root = d_alloc_root(inode);
-
        if (!root) {
                retval = -ENOMEM;
                goto put_back_sb;
@@ -157,24 +156,27 @@ static struct super_block *v9fs_get_sb(struct file_system_type
        stat_result = v9fs_t_stat(v9ses, newfid, &fcall);
        if (stat_result < 0) {
                dprintk(DEBUG_ERROR, "stat error\n");
+               kfree(fcall);
                v9fs_t_clunk(v9ses, newfid);
-               v9fs_put_idpool(newfid, &v9ses->fidpool);
        } else {
                /* Setup the Root Inode */
-               root_fid = v9fs_fid_create(root, v9ses, newfid, 0);
+               kfree(fcall);
+               root_fid = v9fs_fid_create(v9ses, newfid);
                if (root_fid == NULL) {
                        retval = -ENOMEM;
                        goto put_back_sb;
                }
 
+               retval = v9fs_fid_insert(root_fid, root);
+               if (retval < 0)
+                       goto put_back_sb;
+
                root_fid->qid = fcall->params.rstat.stat.qid;
                root->d_inode->i_ino =
                    v9fs_qid2ino(&fcall->params.rstat.stat.qid);
                v9fs_stat2inode(&fcall->params.rstat.stat, root->d_inode, sb);
        }
 
-       kfree(fcall);
-
        if (stat_result < 0) {
                retval = stat_result;
                goto put_back_sb;