| Message ID | 20210830044621.12544-1-ghe@suse.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v2] ocfs2: ocfs2_downconvert_lock failure results in deadlock | expand |
On 8/30/21 12:46 PM, Gang He wrote: > Usually, ocfs2_downconvert_lock() function always downconverts > dlm lock to the expected level for satisfy dlm bast requests > from the other nodes. > But there is a rare situation. When dlm lock conversion is being > canceled, ocfs2_downconvert_lock() function will return -EBUSY. > You need to be aware that ocfs2_cancel_convert() function is > asynchronous in fsdlm implementation. > If we does not requeue this lockres entry, ocfs2 downconvert > thread no longer handles this dlm lock bast request. Then, the > other nodes will not get the dlm lock again, the current node's > process will be blocked when acquire this dlm lock again. > > Signed-off-by: Gang He <ghe@suse.com> Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com> > --- > v2: keep error print in ocfs2_downconvert_lock function, add more detailed > comments in code, add msleep to avoid repeated attempts. > --- > fs/ocfs2/dlmglue.c | 12 ++++++++++++ > 1 file changed, 12 insertions(+) > > diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c > index 48fd369c29a4..f8f561850470 100644 > --- a/fs/ocfs2/dlmglue.c > +++ b/fs/ocfs2/dlmglue.c > @@ -16,6 +16,7 @@ > #include <linux/debugfs.h> > #include <linux/seq_file.h> > #include <linux/time.h> > +#include <linux/delay.h> > #include <linux/quotaops.h> > #include <linux/sched/signal.h> > > @@ -3912,6 +3913,17 @@ static int ocfs2_unblock_lock(struct ocfs2_super *osb, > spin_unlock_irqrestore(&lockres->l_lock, flags); > ret = ocfs2_downconvert_lock(osb, lockres, new_level, set_lvb, > gen); > + /* The dlm lock convert is being cancelled in background, > + * ocfs2_cancel_convert() is asynchronous in fs/dlm, > + * requeue it, try again later. > + */ > + if (ret == -EBUSY) { > + ctl->requeue = 1; > + mlog(ML_BASTS, "lockres %s, ReQ: Downconvert busy\n", > + lockres->l_name); > + ret = 0; > + msleep(20); > + } > > leave: > if (ret) >
diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c index 48fd369c29a4..f8f561850470 100644 --- a/fs/ocfs2/dlmglue.c +++ b/fs/ocfs2/dlmglue.c @@ -16,6 +16,7 @@ #include <linux/debugfs.h> #include <linux/seq_file.h> #include <linux/time.h> +#include <linux/delay.h> #include <linux/quotaops.h> #include <linux/sched/signal.h> @@ -3912,6 +3913,17 @@ static int ocfs2_unblock_lock(struct ocfs2_super *osb, spin_unlock_irqrestore(&lockres->l_lock, flags); ret = ocfs2_downconvert_lock(osb, lockres, new_level, set_lvb, gen); + /* The dlm lock convert is being cancelled in background, + * ocfs2_cancel_convert() is asynchronous in fs/dlm, + * requeue it, try again later. + */ + if (ret == -EBUSY) { + ctl->requeue = 1; + mlog(ML_BASTS, "lockres %s, ReQ: Downconvert busy\n", + lockres->l_name); + ret = 0; + msleep(20); + } leave: if (ret)
Usually, ocfs2_downconvert_lock() function always downconverts dlm lock to the expected level for satisfy dlm bast requests from the other nodes. But there is a rare situation. When dlm lock conversion is being canceled, ocfs2_downconvert_lock() function will return -EBUSY. You need to be aware that ocfs2_cancel_convert() function is asynchronous in fsdlm implementation. If we does not requeue this lockres entry, ocfs2 downconvert thread no longer handles this dlm lock bast request. Then, the other nodes will not get the dlm lock again, the current node's process will be blocked when acquire this dlm lock again. Signed-off-by: Gang He <ghe@suse.com> --- v2: keep error print in ocfs2_downconvert_lock function, add more detailed comments in code, add msleep to avoid repeated attempts. --- fs/ocfs2/dlmglue.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)