OpenVZ Forum: Devel » [RFC][PATCH 0/15] Pid namespaces

Home » Mailing lists » Devel » [RFC][PATCH 0/15] Pid namespaces

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Re: [PATCH 11/15] Signal semantics [message #15379 is a reply to message #15305]

Sun, 29 July 2007 11:23

Oleg Nesterov
Messages: 143
Registered: August 2006

Senior Member

On 07/27, sukadev@us.ibm.com wrote:
>
> Pavel Emelianov [xemul@openvz.org] wrote:
> | Oleg Nesterov wrote:
> | >>
> | >>@@ -1852,7 +1950,7 @@ relock:
> | >> * within that pid space. It can of course get signals from
> | >> * its parent pid space.
> | >> */
> | >>- if (current == task_child_reaper(current))
> | >>+ if (kinfo.flags & KERN_SIGINFO_CINIT)
> | >> continue;
> | >
> | >I think the whole idea is broken, it assumes the sender put something into
> | >"struct sigqueue".
> |
> | Yup. That's the problem. It seems to me that the only way to handle init's
> | signals is to check for permissions in the sending path.
>
> We can check permissions in the sending path - and in fact we do check for
> SIGKILL case (deny_signal_to_container_init() below).
>
> But the receiver knows/decides whether or not the signal is wanted/not. No ?

I can't understand your question. Yes, this is what we are doing currently,
but this is broken by this patch.

> Are you saying we should check/special case all fatal signals ?
>
> |
> | >Suppose that /sbin/init has no handler for (say) SIGTERM, and we send this
> | >signal from the same namespace. send_signal() sets SIGQUEUE_CINIT, but it
> | >is lost because __group_complete_signal() silently "converts" sig_fatal()
> | >signals to SIGKILL using sigaddset().
>
> Yes, I should have called it out, but this patch currently assumes /sbin/init
> (or container-init) has a handler for the fatal signals like SIGTERM

Changelog says nothing about that. And in that case we don't need any complications
except a) deny_signal_to_container_init() (should be named deny_SIGKILL_to_container_init)
and b) "cross-namespace signals must have si_code == SI_KERNEL".

I don't know whether this limitation (/sbin/init must install the handler
for each fatal signal) acceptable or not.

However, we should also take care about sig_kernel_stop() signals, and please
note that it is not possible to install a handler for SIGSTOP.

> | >>+static void encode_sender_info(struct task_struct *t, struct sigqueue *q)
> | >>+{
> | >>+ if (pid_ns_equal(t)) {
> | >>+ if (is_container_init(t)) {
> | >>+ q->flags |= SIGQUEUE_CINIT;
> | >
> | >Ironically, this change carefully preserves the bug we already have :)
> | >
> | >This doesn't protect init from "bad" signal if we send it to sub-thread
> | >of init. Actually, this make the behaviour a bit worse compared to what
> | >we currently have. Currently, at least the main init's thread survives
> | >if we send SIGKILL to sub-thread.
>
> Do you mean "init's main thread" ?

Yes.

> But doesn't SIGKILL to any thread kill
> the entire process ?

It should, but it doesn't if it was sent to init's sub-thread, exactly
because of child_reaper() check in get_signal_to_deliver().

> | >>+ error = deny_signal_to_container_init(t, sig);
> | >>+ if (error)
> | >>+ return error;
> | >
> | >Hm. Could you explain this change? Why do we need a special check for
> | >SIGKILL?
>
> As you pointed out above, SIGKILL goes through the __group_complete_signal()/
> sigaddset() path and bypasses/loses the KERN_SIGINFO_CINIT flag. Other
> sig_fatal() signals take this path too, but we assume for now, container-init
> has a handler.

No, SIGKILL doesn't bypasses send_signal(). IOW, if other parts of this patch
were correct, we don't need this change. If init has a handler, we don't neeed
other parts.

> | >(What about ptrace_attach() btw? If it is possible to send a signal to init
> | > from the "parent" namespace, perhaps it makes sense to allow ptracing as
> | > well).
> |
> | ptracing of tasks fro different namespaces is not possible at all, since
> | strace utility determines the fork()-ed child pid from the parent's eax
> | register, which would contain the pid value as this parent sees his child.
> | But if the strace is in different namespace - it won't be able to find
> | this child with the pid value from parent's eax.
> |
> | Maybe it's worth disabling cross-namespaces ptracing...
>
> I think so too. Its probably not a serious limitation ?

My question was not clear, sorry. And I was confused because I had a false
impression that ptrace_attach() was already changed to use is_container_init().

Afaics, the cross-namespaces ptracing should work (modulo fork() problem
pointed out by Pavel), and probably it is useful.

But we should fix ptrace_attach(), it should not be possible to do PTRACE_ATTACH
to /sbin/init from the _same_ namespace.

Oleg.

Report message to a moderator

[Message index]

		[RFC][PATCH 0/15] Pid namespaces By: Pavel Emelianov on Thu, 26 July 2007 14:45
		[PATCH 1/15] Move exit_task_namespaces() By: Pavel Emelianov on Thu, 26 July 2007 14:46
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Oleg Nesterov on Thu, 26 July 2007 16:47
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Oleg Nesterov on Fri, 27 July 2007 08:07
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Pavel Emelianov on Fri, 27 July 2007 08:24
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Oleg Nesterov on Fri, 27 July 2007 08:35
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Pavel Emelianov on Fri, 27 July 2007 08:37
		Re: [PATCH 1/15] Move exit_task_namespaces() By: dev on Thu, 26 July 2007 16:59
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Dave Hansen on Thu, 26 July 2007 16:10
		Re: [PATCH 1/15] Move exit_task_namespaces() By: xemul on Fri, 27 July 2007 06:38
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Oleg Nesterov on Thu, 02 August 2007 16:18
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Pavel Emelianov on Mon, 06 August 2007 08:00
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Oleg Nesterov on Mon, 06 August 2007 09:54
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Pavel Emelianov on Mon, 06 August 2007 09:58
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Oleg Nesterov on Mon, 06 August 2007 10:38
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Pavel Emelianov on Mon, 06 August 2007 11:21
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Oleg Nesterov on Mon, 06 August 2007 12:52
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Pavel Emelianov on Mon, 06 August 2007 13:38
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Pavel Emelianov on Mon, 06 August 2007 11:29
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Oleg Nesterov on Mon, 06 August 2007 12:48
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Pavel Emelianov on Mon, 06 August 2007 13:36
		Re: [PATCH 1/15] Move exit_task_namespaces() By: Oleg Nesterov on Mon, 06 August 2007 13:55
		[PATCH 2/15] Introduce MS_KERNMOUNT flag By: Pavel Emelianov on Thu, 26 July 2007 14:47
		[PATCH 3/15] kern_siginfo helper By: Pavel Emelianov on Thu, 26 July 2007 14:48
		Re: [PATCH 3/15] kern_siginfo helper By: Oleg Nesterov on Sun, 29 July 2007 11:40
		Re: [PATCH 3/15] kern_siginfo helper By: Pavel Emelianov on Mon, 30 July 2007 06:07
		Re: [PATCH 3/15] kern_siginfo helper By: Sukadev Bhattiprolu on Tue, 31 July 2007 00:21
		[PATCH 4/15] Make proc_flust_task() flush entries from multiple proc trees By: Pavel Emelianov on Thu, 26 July 2007 14:48
		[PATCH 5/15] Introduce struct upid By: Pavel Emelianov on Thu, 26 July 2007 14:49
		Re: [PATCH 5/15] Introduce struct upid By: Oleg Nesterov on Sun, 29 July 2007 09:50
		Re: [PATCH 5/15] Introduce struct upid By: Pavel Emelianov on Mon, 30 July 2007 05:58
		[PATCH 6/15] Make alloc_pid(), free_pid() and put_pid() work with struct upid By: Pavel Emelianov on Thu, 26 July 2007 14:50
		Re: [PATCH 6/15] Make alloc_pid(), free_pid() and put_pid() work with struct upid By: Oleg Nesterov on Sun, 29 July 2007 10:14
		Re: [PATCH 6/15] Make alloc_pid(), free_pid() and put_pid() work with struct upid By: Pavel Emelianov on Mon, 30 July 2007 06:03
		[PATCH 7/15] Helpers to obtain pid numbers By: Pavel Emelianov on Thu, 26 July 2007 14:51
		Re: [PATCH 7/15] Helpers to obtain pid numbers By: Dave Hansen on Thu, 26 July 2007 19:03
		Re: [PATCH 7/15] Helpers to obtain pid numbers By: Pavel Emelianov on Fri, 27 July 2007 06:40
		Re: [PATCH 7/15] Helpers to obtain pid numbers By: Oleg Nesterov on Sun, 29 July 2007 12:08
		Re: [PATCH 7/15] Helpers to obtain pid numbers By: Pavel Emelianov on Mon, 30 July 2007 06:11
		[PATCH 8/15] Helpers to find the task by its numerical ids By: Pavel Emelianov on Thu, 26 July 2007 14:51
		Re: [PATCH 8/15] Helpers to find the task by its numerical ids By: Dave Hansen on Thu, 26 July 2007 19:05
		Re: [PATCH 8/15] Helpers to find the task by its numerical ids By: Pavel Emelianov on Fri, 27 July 2007 06:43
		Re: [PATCH 8/15] Helpers to find the task by its numerical ids By: Oleg Nesterov on Sun, 29 July 2007 12:39
		Re: [PATCH 8/15] Helpers to find the task by its numerical ids By: Pavel Emelianov on Mon, 30 July 2007 06:15
		[PATCH 9/15] Move alloc_pid() after the namespace is cloned By: Pavel Emelianov on Thu, 26 July 2007 14:52
		Re: [PATCH 9/15] Move alloc_pid() after the namespace is cloned By: Oleg Nesterov on Fri, 27 July 2007 15:10
		Re: [PATCH 9/15] Move alloc_pid() after the namespace is cloned By: Pavel Emelianov on Mon, 30 July 2007 06:17
		Re: [PATCH 9/15] Move alloc_pid() after the namespace is cloned By: Sukadev Bhattiprolu on Mon, 30 July 2007 23:43
		Re: [PATCH 9/15] Move alloc_pid() after the namespace is cloned By: Sukadev Bhattiprolu on Tue, 31 July 2007 05:49
		[PATCH 10/15] Make each namespace has its own proc tree By: Pavel Emelianov on Thu, 26 July 2007 14:54
		Re: [PATCH 10/15] Make each namespace has its own proc tree By: Oleg Nesterov on Sun, 29 July 2007 15:56
		Re: [PATCH 10/15] Make each namespace has its own proc tree By: Oleg Nesterov on Sun, 29 July 2007 17:02
		Re: [PATCH 10/15] Make each namespace has its own proc tree By: Pavel Emelianov on Mon, 30 July 2007 06:45
		Re: [PATCH 10/15] Make each namespace has its own proc tree By: Pavel Emelianov on Mon, 30 July 2007 06:43
		[PATCH 11/15] Signal semantics By: Pavel Emelianov on Thu, 26 July 2007 14:55
		Re: [PATCH 11/15] Signal semantics By: Oleg Nesterov on Fri, 27 July 2007 12:30
		Re: [PATCH 11/15] Signal semantics By: Pavel Emelianov on Fri, 27 July 2007 13:38
		Re: [PATCH 11/15] Signal semantics By: serue on Fri, 27 July 2007 19:59
		Re: [PATCH 11/15] Signal semantics By: Sukadev Bhattiprolu on Fri, 27 July 2007 20:23
		Re: [PATCH 11/15] Signal semantics By: Pavel Emelianov on Mon, 30 July 2007 09:34
		Re: [PATCH 11/15] Signal semantics By: Pavel Emelianov on Mon, 30 July 2007 09:31
		Re: [PATCH 11/15] Signal semantics By: serue on Wed, 01 August 2007 16:13
		Re: [PATCH 11/15] Signal semantics By: dev on Thu, 02 August 2007 08:35
		Re: [PATCH 11/15] Signal semantics By: serue on Thu, 02 August 2007 20:09
		Re: [PATCH 11/15] Signal semantics By: Oleg Nesterov on Sun, 29 July 2007 11:23
		[PATCH 12/15] Miscelaneous stuff for pid namespaces By: Pavel Emelianov on Thu, 26 July 2007 14:56
		Re: [PATCH 12/15] Miscelaneous stuff for pid namespaces By: Pavel Emelianov on Fri, 27 July 2007 06:53
		Re: [PATCH 12/15] Miscelaneous stuff for pid namespaces By: Sukadev Bhattiprolu on Fri, 27 July 2007 06:22
		[PATCH 13/15] Clone the pid namespace By: Pavel Emelianov on Thu, 26 July 2007 14:56
		[PATCH 14/15] Destroy pid namespace on init's death By: Pavel Emelianov on Thu, 26 July 2007 14:57
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Oleg Nesterov on Sun, 29 July 2007 10:39
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Pavel Emelianov on Mon, 30 July 2007 11:56
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Oleg Nesterov on Mon, 30 July 2007 15:44
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Oleg Nesterov on Tue, 31 July 2007 09:07
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Sukadev Bhattiprolu on Wed, 01 August 2007 06:16
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Dave Hansen on Wed, 01 August 2007 16:00
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Oleg Nesterov on Wed, 01 August 2007 19:51
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: dev on Thu, 02 August 2007 08:37
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Oleg Nesterov on Thu, 02 August 2007 16:08
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Oleg Nesterov on Thu, 02 August 2007 17:08
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Sukadev Bhattiprolu on Fri, 03 August 2007 06:22
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Oleg Nesterov on Fri, 03 August 2007 10:55
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Sukadev Bhattiprolu on Thu, 02 August 2007 07:37
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Oleg Nesterov on Wed, 01 August 2007 19:48
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Sukadev Bhattiprolu on Thu, 02 August 2007 07:29
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Oleg Nesterov on Thu, 02 August 2007 15:39
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Sukadev Bhattiprolu on Thu, 02 August 2007 17:20
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Oleg Nesterov on Thu, 02 August 2007 17:29
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Sukadev Bhattiprolu on Thu, 02 August 2007 18:36
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Oleg Nesterov on Thu, 02 August 2007 18:49
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: serue on Thu, 02 August 2007 19:13
		Re: [PATCH 14/15] Destroy pid namespace on init's death By: Sukadev Bhattiprolu on Fri, 03 August 2007 21:36
		[PATCH 15/15] Hooks over the code to show correct values to user By: Pavel Emelianov on Thu, 26 July 2007 14:58
		Re: [PATCH 15/15] Hooks over the code to show correct values to user By: Sukadev Bhattiprolu on Fri, 27 July 2007 05:57
		Re: [PATCH 15/15] Hooks over the code to show correct values to user By: Pavel Emelianov on Fri, 27 July 2007 06:44
		Re: [PATCH 15/15] Hooks over the code to show correct values to user By: Oleg Nesterov on Sun, 29 July 2007 14:29
		Re: [PATCH 15/15] Hooks over the code to show correct values to user By: Pavel Emelianov on Mon, 30 July 2007 06:49
		Re: [PATCH 15/15] Hooks over the code to show correct values to user By: Oleg Nesterov on Tue, 31 July 2007 10:04
		Re: [RFC][PATCH 0/15] Pid namespaces By: Sukadev Bhattiprolu on Fri, 27 July 2007 04:22
		Re: [RFC][PATCH 0/15] Pid namespaces By: Sukadev Bhattiprolu on Fri, 27 July 2007 06:08
		Re: [RFC][PATCH 0/15] Pid namespaces By: Pavel Emelianov on Fri, 27 July 2007 06:47

Previous Topic:	Re: [RFC, PATCH] handle the multi-threaded init's exit() properly
Next Topic:	[PATCH 0/14] sysfs cleanups

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Sat Sep 06 12:23:44 GMT 2025

Total time taken to generate the page: 0.15042 seconds