[GIT PULL] ptrace changes for v3.1

[GIT PULL] ptrace changes for v3.1

[Oleg Nesterov]

Linus,

please pull from

	git://git.kernel.org/pub/scm/linux/kernel/git/oleg/misc.git ptrace

More ptrace/jobctl fixes plus PTRACE_SEIZE/PTRACE_LISTEN functionality
from Tejun.

Note:

	- f701e5b7 "connector: add an event for monitoring process tracers" is
	  a bit off-topic, it was routed via ptrace branch because it textually
	  depends on the changes in tracehook/ptrace_attach code.

	- a288eecc "ptrace: kill trivial tracehooks" has the simple conflicts
	  with 248bed4b "[S390] use siginfo for sigtrap signals" pending in s390
	  tree, hopefully this is trivial to resolve.

	  Assuming you didn't pulled s390 yet: s390 version supercedes, only

	  	- tracehook_consider_fatal_signal
	  	+ current->ptrace

	  chunk in arch/s390/kernel/traps.c should be dropped because this was
	  already done by a288eecc

Denys Vlasenko (1):
      ptrace: make former thread ID available via PTRACE_GETEVENTMSG after PTRACE_EVENT_EXEC stop

Oleg Nesterov (17):
      make do_notify_parent() return bool
      kill tracehook_notify_death()
      __ptrace_detach: avoid task_detached(), check do_notify_parent()
      make do_notify_parent() __must_check, update the callers
      reparent_leader: check EXIT_DEAD instead of task_detached()
      kill task_detached()
      do not change dead_task->exit_signal
      redefine thread_group_leader() as exit_signal >= 0
      ptrace: ptrace_reparented() should check same_thread_group()
      ptrace: kill real_parent_is_ptracer() in in favor of ptrace_reparented()
      ptrace: wait_consider_task: s/same_thread_group/ptrace_reparented/
      has_stopped_jobs: s/task_is_stopped/SIGNAL_STOP_STOPPED/
      ptrace_init_task: initialize child->jobctl explicitly
      ptrace: mv send-SIGSTOP from do_fork() to ptrace_init_task()
      ptrace: dont send SIGSTOP on auto-attach if PT_SEIZED
      ptrace: fix ptrace_signal() && STOP_DEQUEUED interaction
      ptrace: do_wait(traced_leader_killed_by_mt_exec) can block forever

Tejun Heo (20):
      ptrace: remove silly wait_trap variable from ptrace_attach()
      job control: rename signal->group_stop and flags to jobctl and update them
      ptrace: ptrace_check_attach(): rename @kill to @ignore_state and add comments
      ptrace: relocate set_current_state(TASK_TRACED) in ptrace_stop()
      job control: introduce JOBCTL_PENDING_MASK and task_clear_jobctl_pending()
      job control: make task_clear_jobctl_pending() clear TRAPPING automatically
      job control: introduce task_set_jobctl_pending()
      ptrace: use bit_waitqueue for TRAPPING instead of wait_chldexit
      signal: remove three noop tracehooks
      job control: introduce JOBCTL_TRAP_STOP and use it for group stop trap
      ptrace: implement PTRACE_SEIZE
      ptrace: implement PTRACE_INTERRUPT
      ptrace: implement TRAP_NOTIFY and use it for group stop events
      ptrace: implement PTRACE_LISTEN
      ptrace: kill task_ptrace()
      ptrace: introduce ptrace_event_enabled() and simplify ptrace_event() and tracehook_prepare_clone()
      ptrace: move SIGTRAP on exec(2) logic to ptrace_event()
      ptrace: kill trivial tracehooks
      ptrace: kill clone/exec tracehooks
      ptrace: s/tracehook_tracer_task()/ptrace_parent()/

Vladimir Zapolskiy (1):
      connector: add an event for monitoring process tracers

 arch/s390/kernel/traps.c    |    4 +-
 drivers/connector/cn_proc.c |   35 ++++
 fs/exec.c                   |   27 +++-
 fs/proc/array.c             |    2 +-
 fs/proc/base.c              |    2 +-
 include/linux/cn_proc.h     |   13 ++
 include/linux/ptrace.h      |  104 ++++++++---
 include/linux/sched.h       |   52 ++++--
 include/linux/tracehook.h   |  385 +--------------------------------------
 kernel/exit.c               |   91 ++++-----
 kernel/fork.c               |   33 +++-
 kernel/ptrace.c             |  197 ++++++++++++++++----
 kernel/signal.c             |  425 +++++++++++++++++++++++++++----------------
 mm/nommu.c                  |    3 +-
 mm/oom_kill.c               |    3 +-
 security/apparmor/domain.c  |    2 +-
 security/selinux/hooks.c    |    4 +-
 17 files changed, 674 insertions(+), 708 deletions(-)

Re: [GIT PULL] ptrace changes for v3.1

[Christoph Hellwig]

On Fri, Jul 22, 2011 at 07:32:37PM +0200, Oleg Nesterov wrote:
> 	- f701e5b7 "connector: add an event for monitoring process tracers" is
> 	  a bit off-topic, it was routed via ptrace branch because it textually
> 	  depends on the changes in tracehook/ptrace_attach code.

Please do not actually add anything to the connector code.  It's a
horrible interface and users should migrate away from it.

Re: [GIT PULL] ptrace changes for v3.1

[Randy Dunlap]

On Fri, 22 Jul 2011 14:03:12 -0400 Christoph Hellwig wrote:

> On Fri, Jul 22, 2011 at 07:32:37PM +0200, Oleg Nesterov wrote:
> > 	- f701e5b7 "connector: add an event for monitoring process tracers" is
> > 	  a bit off-topic, it was routed via ptrace branch because it textually
> > 	  depends on the changes in tracehook/ptrace_attach code.
> 
> Please do not actually add anything to the connector code.  It's a
> horrible interface and users should migrate away from it.

and migrate to  <what>, please??

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

Re: [GIT PULL] ptrace changes for v3.1

[Oleg Nesterov]

(add Evgeniy)

On 07/22, Christoph Hellwig wrote:
>
> On Fri, Jul 22, 2011 at 07:32:37PM +0200, Oleg Nesterov wrote:
> > 	- f701e5b7 "connector: add an event for monitoring process tracers" is
> > 	  a bit off-topic, it was routed via ptrace branch because it textually
> > 	  depends on the changes in tracehook/ptrace_attach code.
>
> Please do not actually add anything to the connector code.  It's a
> horrible interface and users should migrate away from it.

I never understood who can use this interface and for what.
But it has users.

As for this patch. This feature was acked by maintainers,
I do not think I can argue. Fortunately, the patch adds nothing
without CONFIG_PROC_EVENTS.

Oleg.

Re: [GIT PULL] ptrace changes for v3.1

[Evgeniy Polyakov]

On Fri, Jul 22, 2011 at 08:20:18PM +0200, Oleg Nesterov (oleg@redhat.com) wrote:
> On 07/22, Christoph Hellwig wrote:
> >
> > On Fri, Jul 22, 2011 at 07:32:37PM +0200, Oleg Nesterov wrote:
> > > 	- f701e5b7 "connector: add an event for monitoring process tracers" is
> > > 	  a bit off-topic, it was routed via ptrace branch because it textually
> > > 	  depends on the changes in tracehook/ptrace_attach code.
> >
> > Please do not actually add anything to the connector code.  It's a
> > horrible interface and users should migrate away from it.

Besides usual drugs, what else this is all about?

-- 
	Evgeniy Polyakov