fuse: fix reading past EOF
[pandora-kernel.git] / fs / fuse / file.c
index f79de7c..474968f 100644 (file)
@@ -28,7 +28,9 @@ static int fuse_send_open(struct inode *inode, struct file *file, int isdir,
                return PTR_ERR(req);
 
        memset(&inarg, 0, sizeof(inarg));
-       inarg.flags = file->f_flags & ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
+       inarg.flags = file->f_flags & ~(O_CREAT | O_EXCL | O_NOCTTY);
+       if (!fc->atomic_o_trunc)
+               inarg.flags &= ~O_TRUNC;
        req->in.h.opcode = isdir ? FUSE_OPENDIR : FUSE_OPEN;
        req->in.h.nodeid = get_node_id(inode);
        req->in.numargs = 1;
@@ -53,6 +55,9 @@ struct fuse_file *fuse_file_alloc(void)
                if (!ff->reserved_req) {
                        kfree(ff);
                        ff = NULL;
+               } else {
+                       INIT_LIST_HEAD(&ff->write_entry);
+                       atomic_set(&ff->count, 0);
                }
        }
        return ff;
@@ -64,15 +69,39 @@ void fuse_file_free(struct fuse_file *ff)
        kfree(ff);
 }
 
+static struct fuse_file *fuse_file_get(struct fuse_file *ff)
+{
+       atomic_inc(&ff->count);
+       return ff;
+}
+
+static void fuse_release_end(struct fuse_conn *fc, struct fuse_req *req)
+{
+       dput(req->dentry);
+       mntput(req->vfsmount);
+       fuse_put_request(fc, req);
+}
+
+static void fuse_file_put(struct fuse_file *ff)
+{
+       if (atomic_dec_and_test(&ff->count)) {
+               struct fuse_req *req = ff->reserved_req;
+               struct fuse_conn *fc = get_fuse_conn(req->dentry->d_inode);
+               req->end = fuse_release_end;
+               request_send_background(fc, req);
+               kfree(ff);
+       }
+}
+
 void fuse_finish_open(struct inode *inode, struct file *file,
                      struct fuse_file *ff, struct fuse_open_out *outarg)
 {
        if (outarg->open_flags & FOPEN_DIRECT_IO)
                file->f_op = &fuse_direct_io_file_operations;
        if (!(outarg->open_flags & FOPEN_KEEP_CACHE))
-               invalidate_mapping_pages(inode->i_mapping, 0, -1);
+               invalidate_inode_pages2(inode->i_mapping);
        ff->fh = outarg->fh;
-       file->private_data = ff;
+       file->private_data = fuse_file_get(ff);
 }
 
 int fuse_open_common(struct inode *inode, struct file *file, int isdir)
@@ -89,14 +118,6 @@ int fuse_open_common(struct inode *inode, struct file *file, int isdir)
        if (err)
                return err;
 
-       /* If opening the root node, no lookup has been performed on
-          it, so the attributes must be refreshed */
-       if (get_node_id(inode) == FUSE_ROOT_ID) {
-               err = fuse_do_getattr(inode);
-               if (err)
-                       return err;
-       }
-
        ff = fuse_file_alloc();
        if (!ff)
                return -ENOMEM;
@@ -113,8 +134,7 @@ int fuse_open_common(struct inode *inode, struct file *file, int isdir)
        return err;
 }
 
-struct fuse_req *fuse_release_fill(struct fuse_file *ff, u64 nodeid, int flags,
-                                  int opcode)
+void fuse_release_fill(struct fuse_file *ff, u64 nodeid, int flags, int opcode)
 {
        struct fuse_req *req = ff->reserved_req;
        struct fuse_release_in *inarg = &req->misc.release_in;
@@ -126,9 +146,6 @@ struct fuse_req *fuse_release_fill(struct fuse_file *ff, u64 nodeid, int flags,
        req->in.numargs = 1;
        req->in.args[0].size = sizeof(struct fuse_release_in);
        req->in.args[0].value = inarg;
-       kfree(ff);
-
-       return req;
 }
 
 int fuse_release_common(struct inode *inode, struct file *file, int isdir)
@@ -136,15 +153,23 @@ int fuse_release_common(struct inode *inode, struct file *file, int isdir)
        struct fuse_file *ff = file->private_data;
        if (ff) {
                struct fuse_conn *fc = get_fuse_conn(inode);
-               struct fuse_req *req;
 
-               req = fuse_release_fill(ff, get_node_id(inode), file->f_flags,
-                                       isdir ? FUSE_RELEASEDIR : FUSE_RELEASE);
+               fuse_release_fill(ff, get_node_id(inode), file->f_flags,
+                                 isdir ? FUSE_RELEASEDIR : FUSE_RELEASE);
 
                /* Hold vfsmount and dentry until release is finished */
-               req->vfsmount = mntget(file->f_path.mnt);
-               req->dentry = dget(file->f_path.dentry);
-               request_send_background(fc, req);
+               ff->reserved_req->vfsmount = mntget(file->f_path.mnt);
+               ff->reserved_req->dentry = dget(file->f_path.dentry);
+
+               spin_lock(&fc->lock);
+               list_del(&ff->write_entry);
+               spin_unlock(&fc->lock);
+               /*
+                * Normally this will send the RELEASE request,
+                * however if some asynchronous READ or WRITE requests
+                * are outstanding, the sending will be delayed
+                */
+               fuse_file_put(ff);
        }
 
        /* Return value is ignored by VFS */
@@ -165,7 +190,7 @@ static int fuse_release(struct inode *inode, struct file *file)
  * Scramble the ID space with XTEA, so that the value of the files_struct
  * pointer is not exposed to userspace.
  */
-static u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id)
+u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id)
 {
        u32 *k = fc->scramble_key;
        u64 v = (unsigned long) id;
@@ -264,10 +289,9 @@ static int fuse_fsync(struct file *file, struct dentry *de, int datasync)
        return fuse_fsync_common(file, de, datasync, 0);
 }
 
-void fuse_read_fill(struct fuse_req *req, struct file *file,
+void fuse_read_fill(struct fuse_req *req, struct fuse_file *ff,
                    struct inode *inode, loff_t pos, size_t count, int opcode)
 {
-       struct fuse_file *ff = file->private_data;
        struct fuse_read_in *inarg = &req->misc.read_in;
 
        inarg->fh = ff->fh;
@@ -285,10 +309,19 @@ void fuse_read_fill(struct fuse_req *req, struct file *file,
 }
 
 static size_t fuse_send_read(struct fuse_req *req, struct file *file,
-                            struct inode *inode, loff_t pos, size_t count)
+                            struct inode *inode, loff_t pos, size_t count,
+                            fl_owner_t owner)
 {
        struct fuse_conn *fc = get_fuse_conn(inode);
-       fuse_read_fill(req, file, inode, pos, count, FUSE_READ);
+       struct fuse_file *ff = file->private_data;
+
+       fuse_read_fill(req, ff, inode, pos, count, FUSE_READ);
+       if (owner != NULL) {
+               struct fuse_read_in *inarg = &req->misc.read_in;
+
+               inarg->read_flags |= FUSE_READ_LOCKOWNER;
+               inarg->lock_owner = fuse_lock_owner_id(fc, owner);
+       }
        request_send(fc, req);
        return req->out.args[0].size;
 }
@@ -312,7 +345,8 @@ static int fuse_readpage(struct file *file, struct page *page)
        req->out.page_zeroing = 1;
        req->num_pages = 1;
        req->pages[0] = page;
-       fuse_send_read(req, file, inode, page_offset(page), PAGE_CACHE_SIZE);
+       fuse_send_read(req, file, inode, page_offset(page), PAGE_CACHE_SIZE,
+                      NULL);
        err = req->out.h.error;
        fuse_put_request(fc, req);
        if (!err)
@@ -337,20 +371,21 @@ static void fuse_readpages_end(struct fuse_conn *fc, struct fuse_req *req)
                        SetPageError(page);
                unlock_page(page);
        }
+       if (req->ff)
+               fuse_file_put(req->ff);
        fuse_put_request(fc, req);
 }
 
-static void fuse_send_readpages(struct fuse_req *req, struct file *file,
+static void fuse_send_readpages(struct fuse_req *req, struct fuse_file *ff,
                                struct inode *inode)
 {
        struct fuse_conn *fc = get_fuse_conn(inode);
        loff_t pos = page_offset(req->pages[0]);
        size_t count = req->num_pages << PAGE_CACHE_SHIFT;
        req->out.page_zeroing = 1;
-       fuse_read_fill(req, file, inode, pos, count, FUSE_READ);
+       fuse_read_fill(req, ff, inode, pos, count, FUSE_READ);
        if (fc->async_read) {
-               get_file(file);
-               req->file = file;
+               req->ff = fuse_file_get(ff);
                req->end = fuse_readpages_end;
                request_send_background(fc, req);
        } else {
@@ -359,15 +394,15 @@ static void fuse_send_readpages(struct fuse_req *req, struct file *file,
        }
 }
 
-struct fuse_readpages_data {
+struct fuse_fill_data {
        struct fuse_req *req;
-       struct file *file;
+       struct fuse_file *ff;
        struct inode *inode;
 };
 
 static int fuse_readpages_fill(void *_data, struct page *page)
 {
-       struct fuse_readpages_data *data = _data;
+       struct fuse_fill_data *data = _data;
        struct fuse_req *req = data->req;
        struct inode *inode = data->inode;
        struct fuse_conn *fc = get_fuse_conn(inode);
@@ -376,7 +411,7 @@ static int fuse_readpages_fill(void *_data, struct page *page)
            (req->num_pages == FUSE_MAX_PAGES_PER_REQ ||
             (req->num_pages + 1) * PAGE_CACHE_SIZE > fc->max_read ||
             req->pages[req->num_pages - 1]->index + 1 != page->index)) {
-               fuse_send_readpages(req, data->file, inode);
+               fuse_send_readpages(req, data->ff, inode);
                data->req = req = fuse_get_req(fc);
                if (IS_ERR(req)) {
                        unlock_page(page);
@@ -393,14 +428,14 @@ static int fuse_readpages(struct file *file, struct address_space *mapping,
 {
        struct inode *inode = mapping->host;
        struct fuse_conn *fc = get_fuse_conn(inode);
-       struct fuse_readpages_data data;
+       struct fuse_fill_data data;
        int err;
 
        err = -EIO;
        if (is_bad_inode(inode))
                goto out;
 
-       data.file = file;
+       data.ff = file->private_data;
        data.inode = inode;
        data.req = fuse_get_req(fc);
        err = PTR_ERR(data.req);
@@ -410,7 +445,7 @@ static int fuse_readpages(struct file *file, struct address_space *mapping,
        err = read_cache_pages(mapping, pages, fuse_readpages_fill, &data);
        if (!err) {
                if (data.req->num_pages)
-                       fuse_send_readpages(data.req, file, inode);
+                       fuse_send_readpages(data.req, data.ff, inode);
                else
                        fuse_put_request(fc, data.req);
        }
@@ -418,48 +453,88 @@ out:
        return err;
 }
 
-static size_t fuse_send_write(struct fuse_req *req, struct file *file,
-                             struct inode *inode, loff_t pos, size_t count)
+static ssize_t fuse_file_aio_read(struct kiocb *iocb, const struct iovec *iov,
+                                 unsigned long nr_segs, loff_t pos)
+{
+       struct inode *inode = iocb->ki_filp->f_mapping->host;
+
+       if (pos + iov_length(iov, nr_segs) > i_size_read(inode)) {
+               int err;
+               /*
+                * If trying to read past EOF, make sure the i_size
+                * attribute is up-to-date.
+                */
+               err = fuse_update_attributes(inode, NULL, iocb->ki_filp, NULL);
+               if (err)
+                       return err;
+       }
+
+       return generic_file_aio_read(iocb, iov, nr_segs, pos);
+}
+
+static void fuse_write_fill(struct fuse_req *req, struct fuse_file *ff,
+                           struct inode *inode, loff_t pos, size_t count,
+                           int writepage)
 {
        struct fuse_conn *fc = get_fuse_conn(inode);
-       struct fuse_file *ff = file->private_data;
-       struct fuse_write_in inarg;
-       struct fuse_write_out outarg;
+       struct fuse_write_in *inarg = &req->misc.write.in;
+       struct fuse_write_out *outarg = &req->misc.write.out;
 
-       memset(&inarg, 0, sizeof(struct fuse_write_in));
-       inarg.fh = ff->fh;
-       inarg.offset = pos;
-       inarg.size = count;
+       memset(inarg, 0, sizeof(struct fuse_write_in));
+       inarg->fh = ff->fh;
+       inarg->offset = pos;
+       inarg->size = count;
+       inarg->write_flags = writepage ? FUSE_WRITE_CACHE : 0;
        req->in.h.opcode = FUSE_WRITE;
        req->in.h.nodeid = get_node_id(inode);
        req->in.argpages = 1;
        req->in.numargs = 2;
-       req->in.args[0].size = sizeof(struct fuse_write_in);
-       req->in.args[0].value = &inarg;
+       if (fc->minor < 9)
+               req->in.args[0].size = FUSE_COMPAT_WRITE_IN_SIZE;
+       else
+               req->in.args[0].size = sizeof(struct fuse_write_in);
+       req->in.args[0].value = inarg;
        req->in.args[1].size = count;
        req->out.numargs = 1;
        req->out.args[0].size = sizeof(struct fuse_write_out);
-       req->out.args[0].value = &outarg;
+       req->out.args[0].value = outarg;
+}
+
+static size_t fuse_send_write(struct fuse_req *req, struct file *file,
+                             struct inode *inode, loff_t pos, size_t count,
+                             fl_owner_t owner)
+{
+       struct fuse_conn *fc = get_fuse_conn(inode);
+       fuse_write_fill(req, file->private_data, inode, pos, count, 0);
+       if (owner != NULL) {
+               struct fuse_write_in *inarg = &req->misc.write.in;
+               inarg->write_flags |= FUSE_WRITE_LOCKOWNER;
+               inarg->lock_owner = fuse_lock_owner_id(fc, owner);
+       }
        request_send(fc, req);
-       return outarg.size;
+       return req->misc.write.out.size;
 }
 
-static int fuse_prepare_write(struct file *file, struct page *page,
-                             unsigned offset, unsigned to)
+static int fuse_write_begin(struct file *file, struct address_space *mapping,
+                       loff_t pos, unsigned len, unsigned flags,
+                       struct page **pagep, void **fsdata)
 {
-       /* No op */
+       pgoff_t index = pos >> PAGE_CACHE_SHIFT;
+
+       *pagep = __grab_cache_page(mapping, index);
+       if (!*pagep)
+               return -ENOMEM;
        return 0;
 }
 
-static int fuse_commit_write(struct file *file, struct page *page,
-                            unsigned offset, unsigned to)
+static int fuse_buffered_write(struct file *file, struct inode *inode,
+                              loff_t pos, unsigned count, struct page *page)
 {
        int err;
        size_t nres;
-       unsigned count = to - offset;
-       struct inode *inode = page->mapping->host;
        struct fuse_conn *fc = get_fuse_conn(inode);
-       loff_t pos = page_offset(page) + offset;
+       struct fuse_inode *fi = get_fuse_inode(inode);
+       unsigned offset = pos & (PAGE_CACHE_SIZE - 1);
        struct fuse_req *req;
 
        if (is_bad_inode(inode))
@@ -472,23 +547,39 @@ static int fuse_commit_write(struct file *file, struct page *page,
        req->num_pages = 1;
        req->pages[0] = page;
        req->page_offset = offset;
-       nres = fuse_send_write(req, file, inode, pos, count);
+       nres = fuse_send_write(req, file, inode, pos, count, NULL);
        err = req->out.h.error;
        fuse_put_request(fc, req);
-       if (!err && nres != count)
+       if (!err && !nres)
                err = -EIO;
        if (!err) {
-               pos += count;
+               pos += nres;
                spin_lock(&fc->lock);
+               fi->attr_version = ++fc->attr_version;
                if (pos > inode->i_size)
                        i_size_write(inode, pos);
                spin_unlock(&fc->lock);
 
-               if (offset == 0 && to == PAGE_CACHE_SIZE)
+               if (count == PAGE_CACHE_SIZE)
                        SetPageUptodate(page);
        }
        fuse_invalidate_attr(inode);
-       return err;
+       return err ? err : nres;
+}
+
+static int fuse_write_end(struct file *file, struct address_space *mapping,
+                       loff_t pos, unsigned len, unsigned copied,
+                       struct page *page, void *fsdata)
+{
+       struct inode *inode = mapping->host;
+       int res = 0;
+
+       if (copied)
+               res = fuse_buffered_write(file, inode, pos, copied, page);
+
+       unlock_page(page);
+       page_cache_release(page);
+       return res;
 }
 
 static void fuse_release_user_pages(struct fuse_req *req, int write)
@@ -557,9 +648,11 @@ static ssize_t fuse_direct_io(struct file *file, const char __user *buf,
                nbytes = (req->num_pages << PAGE_SHIFT) - req->page_offset;
                nbytes = min(count, nbytes);
                if (write)
-                       nres = fuse_send_write(req, file, inode, pos, nbytes);
+                       nres = fuse_send_write(req, file, inode, pos, nbytes,
+                                              current->files);
                else
-                       nres = fuse_send_read(req, file, inode, pos, nbytes);
+                       nres = fuse_send_read(req, file, inode, pos, nbytes,
+                                             current->files);
                fuse_release_user_pages(req, !write);
                if (req->out.h.error) {
                        if (!res)
@@ -661,7 +754,8 @@ static int convert_fuse_file_lock(const struct fuse_file_lock *ffl,
 }
 
 static void fuse_lk_fill(struct fuse_req *req, struct file *file,
-                        const struct file_lock *fl, int opcode, pid_t pid)
+                        const struct file_lock *fl, int opcode, pid_t pid,
+                        int flock)
 {
        struct inode *inode = file->f_path.dentry->d_inode;
        struct fuse_conn *fc = get_fuse_conn(inode);
@@ -674,6 +768,8 @@ static void fuse_lk_fill(struct fuse_req *req, struct file *file,
        arg->lk.end = fl->fl_end;
        arg->lk.type = fl->fl_type;
        arg->lk.pid = pid;
+       if (flock)
+               arg->lk_flags |= FUSE_LK_FLOCK;
        req->in.h.opcode = opcode;
        req->in.h.nodeid = get_node_id(inode);
        req->in.numargs = 1;
@@ -693,7 +789,7 @@ static int fuse_getlk(struct file *file, struct file_lock *fl)
        if (IS_ERR(req))
                return PTR_ERR(req);
 
-       fuse_lk_fill(req, file, fl, FUSE_GETLK, 0);
+       fuse_lk_fill(req, file, fl, FUSE_GETLK, 0, 0);
        req->out.numargs = 1;
        req->out.args[0].size = sizeof(outarg);
        req->out.args[0].value = &outarg;
@@ -706,7 +802,7 @@ static int fuse_getlk(struct file *file, struct file_lock *fl)
        return err;
 }
 
-static int fuse_setlk(struct file *file, struct file_lock *fl)
+static int fuse_setlk(struct file *file, struct file_lock *fl, int flock)
 {
        struct inode *inode = file->f_path.dentry->d_inode;
        struct fuse_conn *fc = get_fuse_conn(inode);
@@ -723,7 +819,7 @@ static int fuse_setlk(struct file *file, struct file_lock *fl)
        if (IS_ERR(req))
                return PTR_ERR(req);
 
-       fuse_lk_fill(req, file, fl, opcode, pid);
+       fuse_lk_fill(req, file, fl, opcode, pid, flock);
        request_send(fc, req);
        err = req->out.h.error;
        /* locking is restartable */
@@ -749,8 +845,25 @@ static int fuse_file_lock(struct file *file, int cmd, struct file_lock *fl)
                if (fc->no_lock)
                        err = posix_lock_file_wait(file, fl);
                else
-                       err = fuse_setlk(file, fl);
+                       err = fuse_setlk(file, fl, 0);
+       }
+       return err;
+}
+
+static int fuse_file_flock(struct file *file, int cmd, struct file_lock *fl)
+{
+       struct inode *inode = file->f_path.dentry->d_inode;
+       struct fuse_conn *fc = get_fuse_conn(inode);
+       int err;
+
+       if (fc->no_lock) {
+               err = flock_lock_file_wait(file, fl);
+       } else {
+               /* emulate flock with POSIX locks */
+               fl->fl_owner = (fl_owner_t) file;
+               err = fuse_setlk(file, fl, 1);
        }
+
        return err;
 }
 
@@ -793,7 +906,7 @@ static sector_t fuse_bmap(struct address_space *mapping, sector_t block)
 static const struct file_operations fuse_file_operations = {
        .llseek         = generic_file_llseek,
        .read           = do_sync_read,
-       .aio_read       = generic_file_aio_read,
+       .aio_read       = fuse_file_aio_read,
        .write          = do_sync_write,
        .aio_write      = generic_file_aio_write,
        .mmap           = fuse_file_mmap,
@@ -802,6 +915,7 @@ static const struct file_operations fuse_file_operations = {
        .release        = fuse_release,
        .fsync          = fuse_fsync,
        .lock           = fuse_file_lock,
+       .flock          = fuse_file_flock,
        .splice_read    = generic_file_splice_read,
 };
 
@@ -814,13 +928,14 @@ static const struct file_operations fuse_direct_io_file_operations = {
        .release        = fuse_release,
        .fsync          = fuse_fsync,
        .lock           = fuse_file_lock,
+       .flock          = fuse_file_flock,
        /* no mmap and splice_read */
 };
 
 static const struct address_space_operations fuse_file_aops  = {
        .readpage       = fuse_readpage,
-       .prepare_write  = fuse_prepare_write,
-       .commit_write   = fuse_commit_write,
+       .write_begin    = fuse_write_begin,
+       .write_end      = fuse_write_end,
        .readpages      = fuse_readpages,
        .set_page_dirty = fuse_set_page_dirty,
        .bmap           = fuse_bmap,