From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756893AbeEJIKT convert rfc822-to-8bit (ORCPT ); Thu, 10 May 2018 04:10:19 -0400 Received: from prv1-mh.provo.novell.com ([137.65.248.33]:54315 "EHLO prv1-mh.provo.novell.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756671AbeEJIJu (ORCPT ); Thu, 10 May 2018 04:09:50 -0400 Message-Id: <5AF3F989020000F90001A482@prv1-mh.provo.novell.com> X-Mailer: Novell GroupWise Internet Agent 18.0.0 Date: Thu, 10 May 2018 01:49:29 -0600 From: "Gang He" To: , , , "ge changwei" , "Lei Chen" , Cc: , Subject: Re: [Ocfs2-devel] [PATCH] ocfs2: ocfs2_inode_lock_tracker does not distinguish lock level References: <20180510053230.17217-1-lchen@suse.com> In-Reply-To: <20180510053230.17217-1-lchen@suse.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 8BIT Content-Disposition: inline Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hello Joseph, Changwei and Jun/Alex, Please help to take a look at this patch, since the previous patch really has vulnerability in logic, although our test cases did not hit it. Thanks Gang >>> Larry Chen 2018/5/10 13:32 >>> ocfs2_inode_lock_tracker as a variant of ocfs2_inode_lock, is used to prevent deadlock due to recursive lock acquisition. But this function does not distinguish whether the requested level is EX or PR. If a RP lock has been attained, this function will immediately return success afterwards even an EX lock is requested. But actually the return value does not mean that the process got a EX lock, because ocfs2_inode_lock has not been called. When taking lock levels into account, we face some different situations. 1. no lock is held In this case, just lock the inode and return 0 2. We are holding a lock For this situation, things diverges into several cases wanted holding what to do ex ex see 2.1 below ex pr see 2.2 below pr ex see 2.1 below pr pr see 2.1 below 2.1 lock level that is been held is compatible with the wanted level, so no lock action will be tacken. 2.2 Otherwise, an upgrade is needed, but it is forbidden. Reason why upgrade within a process is forbidden is that lock upgrade may cause dead lock. The following illustrate how it happens. process 1 process 2 ocfs2_inode_lock_tracker(ex=0) <====== ocfs2_inode_lock_tracker(ex=1) ocfs2_inode_lock_tracker(ex=1) Signed-off-by: Larry Chen Reviewed-by: Gang He --- fs/ocfs2/dlmglue.c | 119 +++++++++++++++++++++++++++++++++++++++-------------- fs/ocfs2/dlmglue.h | 1 + 2 files changed, 90 insertions(+), 30 deletions(-) diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c index 97a972efab83..68728de12864 100644 --- a/fs/ocfs2/dlmglue.c +++ b/fs/ocfs2/dlmglue.c @@ -788,35 +788,34 @@ static inline void ocfs2_add_holder(struct ocfs2_lock_res *lockres, spin_unlock(&lockres->l_lock); } -static inline void ocfs2_remove_holder(struct ocfs2_lock_res *lockres, - struct ocfs2_lock_holder *oh) -{ - spin_lock(&lockres->l_lock); - list_del(&oh->oh_list); - spin_unlock(&lockres->l_lock); - - put_pid(oh->oh_owner_pid); -} - -static inline int ocfs2_is_locked_by_me(struct ocfs2_lock_res *lockres) +static struct ocfs2_lock_holder * +ocfs2_pid_holder(struct ocfs2_lock_res *lockres, + struct pid *pid) { struct ocfs2_lock_holder *oh; - struct pid *pid; - /* look in the list of holders for one with the current task as owner */ spin_lock(&lockres->l_lock); - pid = task_pid(current); list_for_each_entry(oh, &lockres->l_holders, oh_list) { if (oh->oh_owner_pid == pid) { spin_unlock(&lockres->l_lock); - return 1; + return oh; } } spin_unlock(&lockres->l_lock); + return NULL; +} - return 0; +static inline void ocfs2_remove_holder(struct ocfs2_lock_res *lockres, + struct ocfs2_lock_holder *oh) +{ + spin_lock(&lockres->l_lock); + list_del(&oh->oh_list); + spin_unlock(&lockres->l_lock); + + put_pid(oh->oh_owner_pid); } + static inline void ocfs2_inc_holders(struct ocfs2_lock_res *lockres, int level) { @@ -2610,34 +2609,93 @@ void ocfs2_inode_unlock(struct inode *inode, * * return < 0 on error, return == 0 if there's no lock holder on the stack * before this call, return == 1 if this call would be a recursive locking. + * return == -1 if this lock attempt will cause an upgrade which is forbidden. + * + * When taking lock levels into account,we face some different situations. + * + * 1. no lock is held + * In this case, just lock the inode as requested and return 0 + * + * 2. We are holding a lock + * For this situation, things diverges into several cases + * + * wanted holding what to do + * ex ex see 2.1 below + * ex pr see 2.2 below + * pr ex see 2.1 below + * pr pr see 2.1 below + * + * 2.1 lock level that is been held is compatible + * with the wanted level, so no lock action will be tacken. + * + * 2.2 Otherwise, an upgrade is needed, but it is forbidden. + * + * Reason why upgrade within a process is forbidden is that + * lock upgrade may cause dead lock. The following illustrates + * how it happens. + * + * thread on node1 thread on node2 + * ocfs2_inode_lock_tracker(ex=0) + * + * <====== ocfs2_inode_lock_tracker(ex=1) + * + * ocfs2_inode_lock_tracker(ex=1) */ int ocfs2_inode_lock_tracker(struct inode *inode, struct buffer_head **ret_bh, int ex, struct ocfs2_lock_holder *oh) { - int status; - int arg_flags = 0, has_locked; + int status = 0; struct ocfs2_lock_res *lockres; + struct ocfs2_lock_holder *tmp_oh; + struct pid *pid = task_pid(current); + lockres = &OCFS2_I(inode)->ip_inode_lockres; - has_locked = ocfs2_is_locked_by_me(lockres); - /* Just get buffer head if the cluster lock has been taken */ - if (has_locked) - arg_flags = OCFS2_META_LOCK_GETBH; + tmp_oh = ocfs2_pid_holder(lockres, pid); - if (likely(!has_locked || ret_bh)) { - status = ocfs2_inode_lock_full(inode, ret_bh, ex, arg_flags); + if (!tmp_oh) { + /* + * This corresponds to the case 1. + * We haven't got any lock before. + */ + status = ocfs2_inode_lock_full(inode, ret_bh, ex, 0); if (status < 0) { if (status != -ENOENT) mlog_errno(status); return status; } - } - if (!has_locked) + + oh->oh_ex = ex; ocfs2_add_holder(lockres, oh); + return 0; + } - return has_locked; + if (unlikely(ex && !tmp_oh->oh_ex)) { + /* + * case 2.2 upgrade may cause dead lock, forbid it. + */ + mlog(ML_ERROR, "Recursive locking is not permitted to " + "upgrade to EX level from PR level.\n"); + dump_stack(); + return -EINVAL; + } + + /* + * case 2.1 OCFS2_META_LOCK_GETBH flag make ocfs2_inode_lock_full. + * ignore the lock level and just update it. + */ + if (ret_bh) { + status = ocfs2_inode_lock_full(inode, ret_bh, ex, + OCFS2_META_LOCK_GETBH); + if (status < 0) { + if (status != -ENOENT) + mlog_errno(status); + return status; + } + } + return tmp_oh ? 1 : 0; } void ocfs2_inode_unlock_tracker(struct inode *inode, @@ -2649,12 +2707,13 @@ void ocfs2_inode_unlock_tracker(struct inode *inode, lockres = &OCFS2_I(inode)->ip_inode_lockres; /* had_lock means that the currect process already takes the cluster - * lock previously. If had_lock is 1, we have nothing to do here, and - * it will get unlocked where we got the lock. + * lock previously. + * If had_lock is 1, we have nothing to do here. + * If had_lock is 0, we will release the lock. */ if (!had_lock) { + ocfs2_inode_unlock(inode, oh->oh_ex); ocfs2_remove_holder(lockres, oh); - ocfs2_inode_unlock(inode, ex); } } diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h index 256e0a9067b8..4ec1c828f6e0 100644 --- a/fs/ocfs2/dlmglue.h +++ b/fs/ocfs2/dlmglue.h @@ -96,6 +96,7 @@ struct ocfs2_trim_fs_info { struct ocfs2_lock_holder { struct list_head oh_list; struct pid *oh_owner_pid; + int oh_ex; }; /* ocfs2_inode_lock_full() 'arg_flags' flags */ -- 2.13.6 _______________________________________________ Ocfs2-devel mailing list Ocfs2-devel@oss.oracle.com https://oss.oracle.com/mailman/listinfo/ocfs2-devel From mboxrd@z Thu Jan 1 00:00:00 1970 From: Gang He Date: Thu, 10 May 2018 01:49:29 -0600 Subject: [Ocfs2-devel] [PATCH] ocfs2: ocfs2_inode_lock_tracker does not distinguish lock level In-Reply-To: <20180510053230.17217-1-lchen@suse.com> References: <20180510053230.17217-1-lchen@suse.com> Message-ID: <5AF3F989020000F90001A482@prv1-mh.provo.novell.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: jlbec@evilplan.org, jiangqi903@gmail.com, piaojun@huawei.com, ge changwei , Lei Chen , mfasheh@versity.com Cc: ocfs2-devel@oss.oracle.com, linux-kernel@vger.kernel.org Hello Joseph, Changwei and Jun/Alex, Please help to take a look at this patch, since the previous patch really has vulnerability in logic, although our test cases did not hit it. Thanks Gang >>> Larry Chen 2018/5/10 13:32 >>> ocfs2_inode_lock_tracker as a variant of ocfs2_inode_lock, is used to prevent deadlock due to recursive lock acquisition. But this function does not distinguish whether the requested level is EX or PR. If a RP lock has been attained, this function will immediately return success afterwards even an EX lock is requested. But actually the return value does not mean that the process got a EX lock, because ocfs2_inode_lock has not been called. When taking lock levels into account, we face some different situations. 1. no lock is held In this case, just lock the inode and return 0 2. We are holding a lock For this situation, things diverges into several cases wanted holding what to do ex ex see 2.1 below ex pr see 2.2 below pr ex see 2.1 below pr pr see 2.1 below 2.1 lock level that is been held is compatible with the wanted level, so no lock action will be tacken. 2.2 Otherwise, an upgrade is needed, but it is forbidden. Reason why upgrade within a process is forbidden is that lock upgrade may cause dead lock. The following illustrate how it happens. process 1 process 2 ocfs2_inode_lock_tracker(ex=0) <====== ocfs2_inode_lock_tracker(ex=1) ocfs2_inode_lock_tracker(ex=1) Signed-off-by: Larry Chen Reviewed-by: Gang He --- fs/ocfs2/dlmglue.c | 119 +++++++++++++++++++++++++++++++++++++++-------------- fs/ocfs2/dlmglue.h | 1 + 2 files changed, 90 insertions(+), 30 deletions(-) diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c index 97a972efab83..68728de12864 100644 --- a/fs/ocfs2/dlmglue.c +++ b/fs/ocfs2/dlmglue.c @@ -788,35 +788,34 @@ static inline void ocfs2_add_holder(struct ocfs2_lock_res *lockres, spin_unlock(&lockres->l_lock); } -static inline void ocfs2_remove_holder(struct ocfs2_lock_res *lockres, - struct ocfs2_lock_holder *oh) -{ - spin_lock(&lockres->l_lock); - list_del(&oh->oh_list); - spin_unlock(&lockres->l_lock); - - put_pid(oh->oh_owner_pid); -} - -static inline int ocfs2_is_locked_by_me(struct ocfs2_lock_res *lockres) +static struct ocfs2_lock_holder * +ocfs2_pid_holder(struct ocfs2_lock_res *lockres, + struct pid *pid) { struct ocfs2_lock_holder *oh; - struct pid *pid; - /* look in the list of holders for one with the current task as owner */ spin_lock(&lockres->l_lock); - pid = task_pid(current); list_for_each_entry(oh, &lockres->l_holders, oh_list) { if (oh->oh_owner_pid == pid) { spin_unlock(&lockres->l_lock); - return 1; + return oh; } } spin_unlock(&lockres->l_lock); + return NULL; +} - return 0; +static inline void ocfs2_remove_holder(struct ocfs2_lock_res *lockres, + struct ocfs2_lock_holder *oh) +{ + spin_lock(&lockres->l_lock); + list_del(&oh->oh_list); + spin_unlock(&lockres->l_lock); + + put_pid(oh->oh_owner_pid); } + static inline void ocfs2_inc_holders(struct ocfs2_lock_res *lockres, int level) { @@ -2610,34 +2609,93 @@ void ocfs2_inode_unlock(struct inode *inode, * * return < 0 on error, return == 0 if there's no lock holder on the stack * before this call, return == 1 if this call would be a recursive locking. + * return == -1 if this lock attempt will cause an upgrade which is forbidden. + * + * When taking lock levels into account,we face some different situations. + * + * 1. no lock is held + * In this case, just lock the inode as requested and return 0 + * + * 2. We are holding a lock + * For this situation, things diverges into several cases + * + * wanted holding what to do + * ex ex see 2.1 below + * ex pr see 2.2 below + * pr ex see 2.1 below + * pr pr see 2.1 below + * + * 2.1 lock level that is been held is compatible + * with the wanted level, so no lock action will be tacken. + * + * 2.2 Otherwise, an upgrade is needed, but it is forbidden. + * + * Reason why upgrade within a process is forbidden is that + * lock upgrade may cause dead lock. The following illustrates + * how it happens. + * + * thread on node1 thread on node2 + * ocfs2_inode_lock_tracker(ex=0) + * + * <====== ocfs2_inode_lock_tracker(ex=1) + * + * ocfs2_inode_lock_tracker(ex=1) */ int ocfs2_inode_lock_tracker(struct inode *inode, struct buffer_head **ret_bh, int ex, struct ocfs2_lock_holder *oh) { - int status; - int arg_flags = 0, has_locked; + int status = 0; struct ocfs2_lock_res *lockres; + struct ocfs2_lock_holder *tmp_oh; + struct pid *pid = task_pid(current); + lockres = &OCFS2_I(inode)->ip_inode_lockres; - has_locked = ocfs2_is_locked_by_me(lockres); - /* Just get buffer head if the cluster lock has been taken */ - if (has_locked) - arg_flags = OCFS2_META_LOCK_GETBH; + tmp_oh = ocfs2_pid_holder(lockres, pid); - if (likely(!has_locked || ret_bh)) { - status = ocfs2_inode_lock_full(inode, ret_bh, ex, arg_flags); + if (!tmp_oh) { + /* + * This corresponds to the case 1. + * We haven't got any lock before. + */ + status = ocfs2_inode_lock_full(inode, ret_bh, ex, 0); if (status < 0) { if (status != -ENOENT) mlog_errno(status); return status; } - } - if (!has_locked) + + oh->oh_ex = ex; ocfs2_add_holder(lockres, oh); + return 0; + } - return has_locked; + if (unlikely(ex && !tmp_oh->oh_ex)) { + /* + * case 2.2 upgrade may cause dead lock, forbid it. + */ + mlog(ML_ERROR, "Recursive locking is not permitted to " + "upgrade to EX level from PR level.\n"); + dump_stack(); + return -EINVAL; + } + + /* + * case 2.1 OCFS2_META_LOCK_GETBH flag make ocfs2_inode_lock_full. + * ignore the lock level and just update it. + */ + if (ret_bh) { + status = ocfs2_inode_lock_full(inode, ret_bh, ex, + OCFS2_META_LOCK_GETBH); + if (status < 0) { + if (status != -ENOENT) + mlog_errno(status); + return status; + } + } + return tmp_oh ? 1 : 0; } void ocfs2_inode_unlock_tracker(struct inode *inode, @@ -2649,12 +2707,13 @@ void ocfs2_inode_unlock_tracker(struct inode *inode, lockres = &OCFS2_I(inode)->ip_inode_lockres; /* had_lock means that the currect process already takes the cluster - * lock previously. If had_lock is 1, we have nothing to do here, and - * it will get unlocked where we got the lock. + * lock previously. + * If had_lock is 1, we have nothing to do here. + * If had_lock is 0, we will release the lock. */ if (!had_lock) { + ocfs2_inode_unlock(inode, oh->oh_ex); ocfs2_remove_holder(lockres, oh); - ocfs2_inode_unlock(inode, ex); } } diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h index 256e0a9067b8..4ec1c828f6e0 100644 --- a/fs/ocfs2/dlmglue.h +++ b/fs/ocfs2/dlmglue.h @@ -96,6 +96,7 @@ struct ocfs2_trim_fs_info { struct ocfs2_lock_holder { struct list_head oh_list; struct pid *oh_owner_pid; + int oh_ex; }; /* ocfs2_inode_lock_full() 'arg_flags' flags */ -- 2.13.6 _______________________________________________ Ocfs2-devel mailing list Ocfs2-devel at oss.oracle.com https://oss.oracle.com/mailman/listinfo/ocfs2-devel