[PATCH] ptrace: clear TIF_SYSCALL_TRACE on ptrace detach

[PATCH] ptrace: clear TIF_SYSCALL_TRACE on ptrace detach

[Ales Novak]

On ptrace_detach(), the TIF_SYSCALL_TRACE in thread->flags of the tracee
is not cleared up. This results in the tracehook_report_syscall_* being
called (though there's no longer a tracer listening to that) upon its
further syscalls.

Example scenario - attach "strace" to a running process and kill it (the
strace) with SIGKILL. You'll see that the syscall trace hooks are still
being called.

Signed-off-by: Ales Novak <alnovak@suse.cz>
Cc: Jiri Kosina <jkosina@suse.cz>
---
 kernel/ptrace.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/kernel/ptrace.c b/kernel/ptrace.c
index 1d3b766..634740f 100644
--- a/kernel/ptrace.c
+++ b/kernel/ptrace.c
@@ -85,6 +85,7 @@ void __ptrace_unlink(struct task_struct *child)
 	task_clear_jobctl_pending(child, JOBCTL_TRAP_MASK);
 	task_clear_jobctl_trapping(child);
 
+	clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
 	/*
 	 * Reinstate JOBCTL_STOP_PENDING if group stop is in effect and
 	 * @child isn't dead.
-- 
2.7.0

Re: [PATCH] ptrace: clear TIF_SYSCALL_TRACE on ptrace detach

[Oleg Nesterov]

On 09/01, Ales Novak wrote:
>
> On ptrace_detach(), the TIF_SYSCALL_TRACE in thread->flags of the tracee
> is not cleared up. This results in the tracehook_report_syscall_* being
> called (though there's no longer a tracer listening to that) upon its
> further syscalls.
>
> Example scenario - attach "strace" to a running process and kill it (the
> strace) with SIGKILL. You'll see that the syscall trace hooks are still
> being called.

We do not really care, if the tracer dies the tracee can be in the
inconsistent state anyway.

> --- a/kernel/ptrace.c
> +++ b/kernel/ptrace.c
> @@ -85,6 +85,7 @@ void __ptrace_unlink(struct task_struct *child)
>  	task_clear_jobctl_pending(child, JOBCTL_TRAP_MASK);
>  	task_clear_jobctl_trapping(child);
>
> +	clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);

OK, probably this change makes sense anyway, but then you should
remove another clear_tsk_thread_flag(TIF_SYSCALL_TRACE) in ptrace_detach?

And I'd suggest to move this clear_tsk_thread_flag() up, before
spin_lock(siglock). Otherwise it looks as if we should clear this flag
with ->siglock held.

Oleg.

Re: [PATCH] ptrace: clear TIF_SYSCALL_TRACE on ptrace detach

[Ales Novak]

[-- Attachment #1: Type: text/plain, Size: 1419 bytes --]

On Thu, Sep 01, 2016 at 06:46:13PM +0200, Oleg Nesterov wrote:
> On 09/01, Ales Novak wrote:
> >
> > On ptrace_detach(), the TIF_SYSCALL_TRACE in thread->flags of the tracee
> > is not cleared up. This results in the tracehook_report_syscall_* being
> > called (though there's no longer a tracer listening to that) upon its
> > further syscalls.
> >
> > Example scenario - attach "strace" to a running process and kill it (the
> > strace) with SIGKILL. You'll see that the syscall trace hooks are still
> > being called.
> 
> We do not really care, if the tracer dies the tracee can be in the
> inconsistent state anyway.
> 
> > --- a/kernel/ptrace.c
> > +++ b/kernel/ptrace.c
> > @@ -85,6 +85,7 @@ void __ptrace_unlink(struct task_struct *child)
> >  	task_clear_jobctl_pending(child, JOBCTL_TRAP_MASK);
> >  	task_clear_jobctl_trapping(child);
> >
> > +	clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
> 
> OK, probably this change makes sense anyway, but then you should
> remove another clear_tsk_thread_flag(TIF_SYSCALL_TRACE) in ptrace_detach?

You're right, sorry for not noticing that. Also my changelog was
misleading, I've tried to fix that.

> 
> And I'd suggest to move this clear_tsk_thread_flag() up, before
> spin_lock(siglock). Otherwise it looks as if we should clear this flag
> with ->siglock held.

Good point also.

> 
> Oleg.
> 

-- 
Ales Novak
SUSE L3

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 473 bytes --]

[PATCH] ptrace: clear TIF_SYSCALL_TRACE on ptrace detach

[Ales Novak]

On __ptrace_detach(), called from do_exit()->exit_notify()->
forget_original_parent()->exit_ptrace(), the TIF_SYSCALL_TRACE in
thread->flags of the tracee is not cleared up. This results in the
tracehook_report_syscall_* being called (though there's no longer a
tracer listening to that) upon its further syscalls.

Example scenario - attach "strace" to a running process and kill it (the
strace) with SIGKILL. You'll see that the syscall trace hooks are still
being called.

The clearing of this flag should be moved from ptrace_detach() to
__ptrace_detach().

Signed-off-by: Ales Novak <alnovak@suse.cz>
Cc: Jiri Kosina <jkosina@suse.cz>
---
 kernel/ptrace.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/ptrace.c b/kernel/ptrace.c
index 1d3b766..2a99027 100644
--- a/kernel/ptrace.c
+++ b/kernel/ptrace.c
@@ -73,6 +73,8 @@ void __ptrace_unlink(struct task_struct *child)
 {
 	BUG_ON(!child->ptrace);
 
+	clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+
 	child->parent = child->real_parent;
 	list_del_init(&child->ptrace_entry);
 
@@ -489,7 +491,6 @@ static int ptrace_detach(struct task_struct *child, unsigned int data)
 
 	/* Architecture-specific hardware disable .. */
 	ptrace_disable(child);
-	clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
 
 	write_lock_irq(&tasklist_lock);
 	/*
-- 
2.7.0

Re: [PATCH] ptrace: clear TIF_SYSCALL_TRACE on ptrace detach

[Oleg Nesterov]

On 09/01, Ales Novak wrote:
>
> On __ptrace_detach(), called from do_exit()->exit_notify()->
> forget_original_parent()->exit_ptrace(), the TIF_SYSCALL_TRACE in
> thread->flags of the tracee is not cleared up. This results in the
> tracehook_report_syscall_* being called (though there's no longer a
> tracer listening to that) upon its further syscalls.
>
> Example scenario - attach "strace" to a running process and kill it (the
> strace) with SIGKILL. You'll see that the syscall trace hooks are still
> being called.

To clarify, this patch can't really fix the problem(s), The tracee can
be killed by SIGTRAP in this case. Again, we do not care too much, a lot
of bad things can happen if the tracer dies.

But I agree, this patch makes the things better for free,

Acked-by: Oleg Nesterov <oleg@redhat.com>