Re: [PATCH 4/4] signalfd: make it group-wide, fix posix-timers scheduling
On 08/13, Oleg Nesterov wrote:
>
> Sorry, the patch is not complete. I forgot about signalfd_poll() and somehow I
> forgot to actually tweak de_thread(). I'll send a new one tomorrow.
>
> Something should be done for 2.6.23, imho.
Slightly tested. This all needs furher cleanups (for example,
__exit_signal()->signalfd_detach_locked() is not needed unless we are
group_leader), but this code be done later.
[PATCH 4/4] signalfd: make it group-wide, fix posix-timers scheduling
With this patch any thread can dequeue its own private signals via signalfd,
even if it was created by another sub-thread.
To do so, we use "current" instead of ctx->tsk if the caller is from the same
thread group. This also fixes the scheduling of posix timers broken by the
previous patch.
If the caller doesn't belong to this thread group, we can't handle __SI_TIMER
case properly anyway. Perhaps we should forbid the cross-process signalfd usage
and convert ctx->tsk to ctx->sighand.
Signed-off-by: Oleg Nesterov <[EMAIL PROTECTED]>
--- t/fs/signalfd.c~4_GROUP 2007-08-09 19:59:27.0 +0400
+++ t/fs/signalfd.c 2007-08-14 21:26:31.0 +0400
@@ -56,12 +56,18 @@ static int signalfd_lock(struct signalfd
sighand = lock_task_sighand(lk->tsk, >flags);
rcu_read_unlock();
- if (sighand && !ctx->tsk) {
+ if (!sighand)
+ return 0;
+
+ if (!ctx->tsk) {
unlock_task_sighand(lk->tsk, >flags);
- sighand = NULL;
+ return 0;
}
- return sighand != NULL;
+ if (lk->tsk->tgid == current->tgid)
+ lk->tsk = current;
+
+ return 1;
}
static void signalfd_unlock(struct signalfd_lockctx *lk)
@@ -331,7 +337,7 @@ asmlinkage long sys_signalfd(int ufd, si
init_waitqueue_head(>wqh);
ctx->sigmask = sigmask;
- ctx->tsk = current;
+ ctx->tsk = current->group_leader;
sighand = current->sighand;
/*
--- t/fs/exec.c~4_GROUP 2007-07-28 16:58:16.0 +0400
+++ t/fs/exec.c 2007-08-13 21:05:34.0 +0400
@@ -780,18 +780,12 @@ static int de_thread(struct task_struct
int count;
/*
-* Tell all the sighand listeners that this sighand has
-* been detached. The signalfd_detach() function grabs the
-* sighand lock, if signal listeners are present on the sighand.
-*/
- signalfd_detach(tsk);
-
- /*
* If we don't share sighandlers, then we aren't sharing anything
* and we can just re-use it all.
*/
if (atomic_read(>count) <= 1) {
BUG_ON(atomic_read(>count) != 1);
+ signalfd_detach(tsk);
exit_itimers(sig);
return 0;
}
@@ -930,6 +924,7 @@ static int de_thread(struct task_struct
sig->flags = 0;
no_thread_group:
+ signalfd_detach(tsk);
exit_itimers(sig);
if (leader)
release_task(leader);
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 4/4] signalfd: make it group-wide, fix posix-timers scheduling
On 08/13, Oleg Nesterov wrote:
Sorry, the patch is not complete. I forgot about signalfd_poll() and somehow I
forgot to actually tweak de_thread(). I'll send a new one tomorrow.
Something should be done for 2.6.23, imho.
Slightly tested. This all needs furher cleanups (for example,
__exit_signal()-signalfd_detach_locked() is not needed unless we are
group_leader), but this code be done later.
[PATCH 4/4] signalfd: make it group-wide, fix posix-timers scheduling
With this patch any thread can dequeue its own private signals via signalfd,
even if it was created by another sub-thread.
To do so, we use current instead of ctx-tsk if the caller is from the same
thread group. This also fixes the scheduling of posix timers broken by the
previous patch.
If the caller doesn't belong to this thread group, we can't handle __SI_TIMER
case properly anyway. Perhaps we should forbid the cross-process signalfd usage
and convert ctx-tsk to ctx-sighand.
Signed-off-by: Oleg Nesterov [EMAIL PROTECTED]
--- t/fs/signalfd.c~4_GROUP 2007-08-09 19:59:27.0 +0400
+++ t/fs/signalfd.c 2007-08-14 21:26:31.0 +0400
@@ -56,12 +56,18 @@ static int signalfd_lock(struct signalfd
sighand = lock_task_sighand(lk-tsk, lk-flags);
rcu_read_unlock();
- if (sighand !ctx-tsk) {
+ if (!sighand)
+ return 0;
+
+ if (!ctx-tsk) {
unlock_task_sighand(lk-tsk, lk-flags);
- sighand = NULL;
+ return 0;
}
- return sighand != NULL;
+ if (lk-tsk-tgid == current-tgid)
+ lk-tsk = current;
+
+ return 1;
}
static void signalfd_unlock(struct signalfd_lockctx *lk)
@@ -331,7 +337,7 @@ asmlinkage long sys_signalfd(int ufd, si
init_waitqueue_head(ctx-wqh);
ctx-sigmask = sigmask;
- ctx-tsk = current;
+ ctx-tsk = current-group_leader;
sighand = current-sighand;
/*
--- t/fs/exec.c~4_GROUP 2007-07-28 16:58:16.0 +0400
+++ t/fs/exec.c 2007-08-13 21:05:34.0 +0400
@@ -780,18 +780,12 @@ static int de_thread(struct task_struct
int count;
/*
-* Tell all the sighand listeners that this sighand has
-* been detached. The signalfd_detach() function grabs the
-* sighand lock, if signal listeners are present on the sighand.
-*/
- signalfd_detach(tsk);
-
- /*
* If we don't share sighandlers, then we aren't sharing anything
* and we can just re-use it all.
*/
if (atomic_read(oldsighand-count) = 1) {
BUG_ON(atomic_read(sig-count) != 1);
+ signalfd_detach(tsk);
exit_itimers(sig);
return 0;
}
@@ -930,6 +924,7 @@ static int de_thread(struct task_struct
sig-flags = 0;
no_thread_group:
+ signalfd_detach(tsk);
exit_itimers(sig);
if (leader)
release_task(leader);
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 4/4] signalfd: make it group-wide, fix posix-timers scheduling
On 08/12, Oleg Nesterov wrote: > > Untested, user visible change. > > With this patch any thread can dequeue its own private signals via signalfd, > even if it was created by another sub-thread. > > To do so, we pass "current" to dequeue_signal() if the caller is from the same > thread group. This also fixes the scheduling of posix timers broken by the > previous patch. > > If the caller doesn't belong to this thread group, we can't handle __SI_TIMER > case properly anyway. Perhaps we should forbid the cross-process signalfd > usage > and convert ctx->tsk to ctx->sighand. Sorry, the patch is not complete. I forgot about signalfd_poll() and somehow I forgot to actually tweak de_thread(). I'll send a new one tomorrow. Something should be done for 2.6.23, imho. Oleg. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 4/4] signalfd: make it group-wide, fix posix-timers scheduling
On 08/12, Oleg Nesterov wrote: Untested, user visible change. With this patch any thread can dequeue its own private signals via signalfd, even if it was created by another sub-thread. To do so, we pass current to dequeue_signal() if the caller is from the same thread group. This also fixes the scheduling of posix timers broken by the previous patch. If the caller doesn't belong to this thread group, we can't handle __SI_TIMER case properly anyway. Perhaps we should forbid the cross-process signalfd usage and convert ctx-tsk to ctx-sighand. Sorry, the patch is not complete. I forgot about signalfd_poll() and somehow I forgot to actually tweak de_thread(). I'll send a new one tomorrow. Something should be done for 2.6.23, imho. Oleg. - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
