lkml.org 
[lkml]   [2008]   [Dec]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 2/3] ftrace: use struct pid

On Thu, 4 Dec 2008, Dave Hansen wrote:
> > >>
> > >> Could we get away with sticking the rcu_read_{un}lock() inside those
> > >> macros? Those are going to get used in pretty high level code and we're
> > >> allowed to nest rcu_read_lock(). No danger of deadlocks or lock
> > >> inversions.
> > >
> > > Why don't any of the other users of do_each_pid_task() use
> > > rcu_read_lock()? They all seem to be under read_lock(&tasklist_lock)
> > > (except one is under a write lock of the same).
> >
> > We probably should. Historically read_lock(&tasklist_lock) implies
> > rcu_read_lock().
>
> You mean because the current task can't go through a quiescent period
> until it hits userspace, and we can't go to userspace while holding
> read_lock()? Nah, that's not subtle. ;)

Has nothing to do with userspace. We can not go through a quiescent period
while holding a rcu_read_lock, or if preemption is disabled. read_lock
prevents preemption, as does spin_locks.

>
> > And the tasklist lock is what we hold when it is safe.
> >
> > But if you look at find_vpid we should be holding just the rcu lock there.
>
> Yup, I see it there.
>
> So, any reason not to do this? Brown-bag compile tested.
>
> Signed-off-by: Dave Hansen <dave@linux.vnet.ibm.com>
>
> ---
>
> linux-2.6.git-dave/include/linux/pid.h | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff -puN include/linux/pid.h~put-rcu-ops-in-do_each_pid_task include/linux/pid.h
> --- linux-2.6.git/include/linux/pid.h~put-rcu-ops-in-do_each_pid_task 2008-12-04 06:03:09.000000000 -0800
> +++ linux-2.6.git-dave/include/linux/pid.h 2008-12-04 06:19:35.000000000 -0800
> @@ -147,6 +147,7 @@ pid_t pid_vnr(struct pid *pid);
> #define do_each_pid_task(pid, type, task) \
> do { \
> struct hlist_node *pos___; \
> + rcu_read_lock(); \
> if (pid != NULL) \
> hlist_for_each_entry_rcu((task), pos___, \
> &pid->tasks[type], pids[type].node) {
> @@ -159,6 +160,7 @@ pid_t pid_vnr(struct pid *pid);
> if (type == PIDTYPE_PID) \
> break; \
> } \
> + rcu_read_unlock(); \
> } while (0)

That probably could work.

-- Steve



\
 
 \ /
  Last update: 2008-12-04 16:37    [from the cache]
©2003-2011 Jasper Spaans