Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1760645Ab1CEBIm (ORCPT ); Fri, 4 Mar 2011 20:08:42 -0500 Received: from kroah.org ([198.145.64.141]:47681 "EHLO coco.kroah.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1760563Ab1CEBFA (ORCPT ); Fri, 4 Mar 2011 20:05:00 -0500 X-Mailbox-Line: From gregkh@clark.kroah.org Fri Mar 4 17:03:52 2011 Message-Id: <20110305010352.308656138@clark.kroah.org> User-Agent: quilt/0.48-11.2 Date: Fri, 04 Mar 2011 17:02:59 -0800 From: Greg KH To: linux-kernel@vger.kernel.org, stable@kernel.org Cc: stable-review@kernel.org, torvalds@linux-foundation.org, akpm@linux-foundation.org, alan@lxorguk.ukuu.org.uk, Miklos Szeredi Subject: [05/18] fuse: fix hang of single threaded fuseblk filesystem In-Reply-To: <20110305010410.GA18668@kroah.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 4363 Lines: 148 2.6.32-longterm review patch. If anyone has any objections, please let us know. ------------------ From: Miklos Szeredi commit 5a18ec176c934ca1bc9dc61580a5e0e90a9b5733 upstream. Single threaded NTFS-3G could get stuck if a delayed RELEASE reply triggered a DESTROY request via path_put(). Fix this by a) making RELEASE requests synchronous, whenever possible, on fuseblk filesystems b) if not possible (triggered by an asynchronous read/write) then do the path_put() in a separate thread with schedule_work(). Reported-by: Oliver Neukum Signed-off-by: Miklos Szeredi Signed-off-by: Greg Kroah-Hartman --- fs/fuse/file.c | 52 +++++++++++++++++++++++++++++++++++++++++++++------- fs/fuse/fuse_i.h | 6 +++++- 2 files changed, 50 insertions(+), 8 deletions(-) --- a/fs/fuse/file.c +++ b/fs/fuse/file.c @@ -86,18 +86,52 @@ struct fuse_file *fuse_file_get(struct f return ff; } +static void fuse_release_async(struct work_struct *work) +{ + struct fuse_req *req; + struct fuse_conn *fc; + struct path path; + + req = container_of(work, struct fuse_req, misc.release.work); + path = req->misc.release.path; + fc = get_fuse_conn(path.dentry->d_inode); + + fuse_put_request(fc, req); + path_put(&path); +} + static void fuse_release_end(struct fuse_conn *fc, struct fuse_req *req) { - path_put(&req->misc.release.path); + if (fc->destroy_req) { + /* + * If this is a fuseblk mount, then it's possible that + * releasing the path will result in releasing the + * super block and sending the DESTROY request. If + * the server is single threaded, this would hang. + * For this reason do the path_put() in a separate + * thread. + */ + atomic_inc(&req->count); + INIT_WORK(&req->misc.release.work, fuse_release_async); + schedule_work(&req->misc.release.work); + } else { + path_put(&req->misc.release.path); + } } -static void fuse_file_put(struct fuse_file *ff) +static void fuse_file_put(struct fuse_file *ff, bool sync) { if (atomic_dec_and_test(&ff->count)) { struct fuse_req *req = ff->reserved_req; - req->end = fuse_release_end; - fuse_request_send_background(ff->fc, req); + if (sync) { + fuse_request_send(ff->fc, req); + path_put(&req->misc.release.path); + fuse_put_request(ff->fc, req); + } else { + req->end = fuse_release_end; + fuse_request_send_background(ff->fc, req); + } kfree(ff); } } @@ -219,8 +253,12 @@ void fuse_release_common(struct file *fi * Normally this will send the RELEASE request, however if * some asynchronous READ or WRITE requests are outstanding, * the sending will be delayed. + * + * Make the release synchronous if this is a fuseblk mount, + * synchronous RELEASE is allowed (and desirable) in this case + * because the server can be trusted not to screw up. */ - fuse_file_put(ff); + fuse_file_put(ff, ff->fc->destroy_req != NULL); } static int fuse_open(struct inode *inode, struct file *file) @@ -549,7 +587,7 @@ static void fuse_readpages_end(struct fu unlock_page(page); } if (req->ff) - fuse_file_put(req->ff); + fuse_file_put(req->ff, false); } static void fuse_send_readpages(struct fuse_req *req, struct file *file) @@ -1129,7 +1167,7 @@ static ssize_t fuse_direct_write(struct static void fuse_writepage_free(struct fuse_conn *fc, struct fuse_req *req) { __free_page(req->pages[0]); - fuse_file_put(req->ff); + fuse_file_put(req->ff, false); } static void fuse_writepage_finish(struct fuse_conn *fc, struct fuse_req *req) --- a/fs/fuse/fuse_i.h +++ b/fs/fuse/fuse_i.h @@ -21,6 +21,7 @@ #include #include #include +#include /** Max number of pages that can be used in a single read request */ #define FUSE_MAX_PAGES_PER_REQ 32 @@ -254,7 +255,10 @@ struct fuse_req { union { struct fuse_forget_in forget_in; struct { - struct fuse_release_in in; + union { + struct fuse_release_in in; + struct work_struct work; + }; struct path path; } release; struct fuse_init_in init_in; -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/