When we have to take both dlm->master_lock and lockres->spinlock,
take them in order
lockres->spinlock and then dlm->master_lock.
The patch fixes a violation of the rule.
We can simply move taking dlm->master_lock to where we have dropped
res->spinlock
since when we access res->state and free mle memory we don't need
master_lock's
protection.
Signed-off-by: Wengang Wang <wen.gang.wang at oracle.com>
---
fs/ocfs2/dlm/dlmmaster.c | 5 ++---
1 files changed, 2 insertions(+), 3 deletions(-)
diff --git a/fs/ocfs2/dlm/dlmmaster.c b/fs/ocfs2/dlm/dlmmaster.c
index 94b97fc..6d098b8 100644
--- a/fs/ocfs2/dlm/dlmmaster.c
+++ b/fs/ocfs2/dlm/dlmmaster.c
@@ -3050,8 +3050,6 @@ int dlm_migrate_request_handler(struct o2net_msg *msg, u32
len, void *data,
/* check for pre-existing lock */
spin_lock(&dlm->spinlock);
res = __dlm_lookup_lockres(dlm, name, namelen, hash);
- spin_lock(&dlm->master_lock);
-
if (res) {
spin_lock(&res->spinlock);
if (res->state & DLM_LOCK_RES_RECOVERING) {
@@ -3069,14 +3067,15 @@ int dlm_migrate_request_handler(struct o2net_msg *msg,
u32 len, void *data,
spin_unlock(&res->spinlock);
}
+ spin_lock(&dlm->master_lock);
/* ignore status. only nonzero status would BUG. */
ret = dlm_add_migration_mle(dlm, res, mle, &oldmle,
name, namelen,
migrate->new_master,
migrate->master);
-unlock:
spin_unlock(&dlm->master_lock);
+unlock:
spin_unlock(&dlm->spinlock);
if (oldmle) {
--
1.6.6.1
Just a reminder. wengang. On 10-07-16 23:13, Wengang Wang wrote:> When we have to take both dlm->master_lock and lockres->spinlock, > take them in order > > lockres->spinlock and then dlm->master_lock. > > The patch fixes a violation of the rule. > We can simply move taking dlm->master_lock to where we have dropped res->spinlock > since when we access res->state and free mle memory we don't need master_lock's > protection. > > Signed-off-by: Wengang Wang <wen.gang.wang at oracle.com> > --- > fs/ocfs2/dlm/dlmmaster.c | 5 ++--- > 1 files changed, 2 insertions(+), 3 deletions(-) > > diff --git a/fs/ocfs2/dlm/dlmmaster.c b/fs/ocfs2/dlm/dlmmaster.c > index 94b97fc..6d098b8 100644 > --- a/fs/ocfs2/dlm/dlmmaster.c > +++ b/fs/ocfs2/dlm/dlmmaster.c > @@ -3050,8 +3050,6 @@ int dlm_migrate_request_handler(struct o2net_msg *msg, u32 len, void *data, > /* check for pre-existing lock */ > spin_lock(&dlm->spinlock); > res = __dlm_lookup_lockres(dlm, name, namelen, hash); > - spin_lock(&dlm->master_lock); > - > if (res) { > spin_lock(&res->spinlock); > if (res->state & DLM_LOCK_RES_RECOVERING) { > @@ -3069,14 +3067,15 @@ int dlm_migrate_request_handler(struct o2net_msg *msg, u32 len, void *data, > spin_unlock(&res->spinlock); > } > > + spin_lock(&dlm->master_lock); > /* ignore status. only nonzero status would BUG. */ > ret = dlm_add_migration_mle(dlm, res, mle, &oldmle, > name, namelen, > migrate->new_master, > migrate->master); > > -unlock: > spin_unlock(&dlm->master_lock); > +unlock: > spin_unlock(&dlm->spinlock); > > if (oldmle) { > -- > 1.6.6.1 > > > _______________________________________________ > Ocfs2-devel mailing list > Ocfs2-devel at oss.oracle.com > http://oss.oracle.com/mailman/listinfo/ocfs2-devel
On Fri, Jul 16, 2010 at 11:13:33PM +0800, Wengang Wang wrote:> When we have to take both dlm->master_lock and lockres->spinlock, > take them in order > > lockres->spinlock and then dlm->master_lock. > > The patch fixes a violation of the rule. > We can simply move taking dlm->master_lock to where we have dropped res->spinlock > since when we access res->state and free mle memory we don't need master_lock's > protection. > > Signed-off-by: Wengang Wang <wen.gang.wang at oracle.com>This patch is now in the fixes branch of ocfs2.git. Joel -- "Three o'clock is always too late or too early for anything you want to do." - Jean-Paul Sartre Joel Becker Consulting Software Developer Oracle E-mail: joel.becker at oracle.com Phone: (650) 506-8127