From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756898Ab0CaFXF (ORCPT ); Wed, 31 Mar 2010 01:23:05 -0400 Received: from sj-iport-6.cisco.com ([171.71.176.117]:36905 "EHLO sj-iport-6.cisco.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755875Ab0CaFXC (ORCPT ); Wed, 31 Mar 2010 01:23:02 -0400 Authentication-Results: sj-iport-6.cisco.com; dkim=neutral (message not signed) header.i=none X-IronPort-Anti-Spam-Filtered: true X-IronPort-Anti-Spam-Result: AvsEAIt5skurR7Ht/2dsb2JhbACbKHGmTpkThQAEgyA X-IronPort-AV: E=Sophos;i="4.51,339,1267401600"; d="scan'208";a="505832657" From: Roland Dreier To: Arnd Bergmann Cc: Frederic Weisbecker , linux-kernel@vger.kernel.org, Matthew Wilcox , Thomas Gleixner , jblunck@suse.de, Alan Cox , Ingo Molnar Subject: Re: [GIT, RFC] Killing the Big Kernel Lock References: <201003242240.54907.arnd@arndb.de> <201003242259.58535.arnd@arndb.de> X-Message-Flag: Warning: May contain useful information Date: Tue, 30 Mar 2010 22:22:57 -0700 In-Reply-To: <201003242259.58535.arnd@arndb.de> (Arnd Bergmann's message of "Wed, 24 Mar 2010 22:59:57 +0100") Message-ID: User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/23.1 (gnu/linux) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org OK, I added the following to my tree, currently queued in my for-next branch for 2.6.35: IB: Explicitly rule out llseek to avoid BKL in default_llseek() Several RDMA user-access drivers have file_operations structures with no .llseek method set. None of the drivers actually do anything with f_pos, so this means llseek is essentially a NOP, instead of returning an error as leaving other file_operations methods unimplemented would do. This is mostly harmless, except that a NULL .llseek means that default_llseek() is used, and this function grabs the BKL, which we would like to avoid. Since llseek does nothing useful on these files, we would like it to return an error to userspace instead of silently grabbing the BKL and succeeding. For nearly all of the file types, we take the belt-and-suspenders approach of setting the .llseek method to no_llseek and also calling nonseekable_open(); the exception is the uverbs_event files, which are created with anon_inode_getfile(), which already sets f_mode the same way as nonseekable_open() would. This work is motivated by Arnd Bergmann's bkl-removal tree. Signed-off-by: Roland Dreier --- drivers/infiniband/core/ucm.c | 3 ++- drivers/infiniband/core/ucma.c | 4 +++- drivers/infiniband/core/user_mad.c | 12 ++++++++---- drivers/infiniband/core/uverbs_main.c | 11 +++++++---- 4 files changed, 20 insertions(+), 10 deletions(-) diff --git a/drivers/infiniband/core/ucm.c b/drivers/infiniband/core/ucm.c index 017d6e2..7ef3954 100644 --- a/drivers/infiniband/core/ucm.c +++ b/drivers/infiniband/core/ucm.c @@ -1180,7 +1180,7 @@ static int ib_ucm_open(struct inode *inode, struct file *filp) file->filp = filp; file->device = container_of(inode->i_cdev, struct ib_ucm_device, cdev); - return 0; + return nonseekable_open(inode, filp); } static int ib_ucm_close(struct inode *inode, struct file *filp) @@ -1228,6 +1228,7 @@ static const struct file_operations ucm_fops = { .release = ib_ucm_close, .write = ib_ucm_write, .poll = ib_ucm_poll, + .llseek = no_llseek, }; static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr, diff --git a/drivers/infiniband/core/ucma.c b/drivers/infiniband/core/ucma.c index b2e16c3..09d4a3b 100644 --- a/drivers/infiniband/core/ucma.c +++ b/drivers/infiniband/core/ucma.c @@ -1219,7 +1219,8 @@ static int ucma_open(struct inode *inode, struct file *filp) filp->private_data = file; file->filp = filp; - return 0; + + return nonseekable_open(inode, filp); } static int ucma_close(struct inode *inode, struct file *filp) @@ -1249,6 +1250,7 @@ static const struct file_operations ucma_fops = { .release = ucma_close, .write = ucma_write, .poll = ucma_poll, + .llseek = no_llseek, }; static struct miscdevice ucma_misc = { diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c index 04b585e..2bb9703 100644 --- a/drivers/infiniband/core/user_mad.c +++ b/drivers/infiniband/core/user_mad.c @@ -780,7 +780,7 @@ static int ib_umad_open(struct inode *inode, struct file *filp) { struct ib_umad_port *port; struct ib_umad_file *file; - int ret = 0; + int ret; port = container_of(inode->i_cdev, struct ib_umad_port, cdev); if (port) @@ -813,6 +813,8 @@ static int ib_umad_open(struct inode *inode, struct file *filp) list_add_tail(&file->port_list, &port->file_list); + ret = nonseekable_open(inode, filp); + out: mutex_unlock(&port->file_mutex); return ret; @@ -865,7 +867,8 @@ static const struct file_operations umad_fops = { .compat_ioctl = ib_umad_compat_ioctl, #endif .open = ib_umad_open, - .release = ib_umad_close + .release = ib_umad_close, + .llseek = no_llseek, }; static int ib_umad_sm_open(struct inode *inode, struct file *filp) @@ -902,7 +905,7 @@ static int ib_umad_sm_open(struct inode *inode, struct file *filp) filp->private_data = port; - return 0; + return nonseekable_open(inode, filp); fail: kref_put(&port->umad_dev->ref, ib_umad_release_dev); @@ -932,7 +935,8 @@ static int ib_umad_sm_close(struct inode *inode, struct file *filp) static const struct file_operations umad_sm_fops = { .owner = THIS_MODULE, .open = ib_umad_sm_open, - .release = ib_umad_sm_close + .release = ib_umad_sm_close, + .llseek = no_llseek, }; static struct ib_client umad_client = { diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index 1444f95..a16a91e 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -385,7 +385,8 @@ static const struct file_operations uverbs_event_fops = { .read = ib_uverbs_event_read, .poll = ib_uverbs_event_poll, .release = ib_uverbs_event_close, - .fasync = ib_uverbs_event_fasync + .fasync = ib_uverbs_event_fasync, + .llseek = no_llseek, }; void ib_uverbs_comp_handler(struct ib_cq *cq, void *cq_context) @@ -639,7 +640,7 @@ static int ib_uverbs_open(struct inode *inode, struct file *filp) filp->private_data = file; - return 0; + return nonseekable_open(inode, filp); err_module: module_put(dev->ib_dev->owner); @@ -667,7 +668,8 @@ static const struct file_operations uverbs_fops = { .owner = THIS_MODULE, .write = ib_uverbs_write, .open = ib_uverbs_open, - .release = ib_uverbs_close + .release = ib_uverbs_close, + .llseek = no_llseek, }; static const struct file_operations uverbs_mmap_fops = { @@ -675,7 +677,8 @@ static const struct file_operations uverbs_mmap_fops = { .write = ib_uverbs_write, .mmap = ib_uverbs_mmap, .open = ib_uverbs_open, - .release = ib_uverbs_close + .release = ib_uverbs_close, + .llseek = no_llseek, }; static struct ib_client uverbs_client = { -- 1.7.0.3 -- Roland Dreier || For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/index.html