From: Oleg Nesterov <oleg@redhat.com>
To: Linus Torvalds <torvalds@linux-foundation.org>,
Roland McGrath <roland@hack.frob.com>, Tejun Heo <tj@kernel.org>
Cc: Denys Vlasenko <dvlasenk@redhat.com>,
KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>,
Matt Fleming <matt.fleming@linux.intel.com>,
linux-kernel@vger.kernel.org
Subject: [PATCH 7/8] vfork: make it stoppable/traceable
Date: Wed, 27 Jul 2011 18:34:12 +0200 [thread overview]
Message-ID: <20110727163412.GG23793@redhat.com> (raw)
In-Reply-To: <20110727163159.GA23785@redhat.com>
Make vfork() stoppable/traceable.
Change clone_vfork_finish() paths to block all signals except
SIGKILL | SIGSTOP and do wait_for_completion_interruptible().
This means we should restart after the stop/ptrace_attach or
the spurious wakeup, implement clone_vfork_restart().
-ERESTART_RESTARTBLOCK is safe, we can never dequeue a signal
which has a handler, thus we can never return to the user-space.
Unless the debugger changes regs of course, but this is fine.
Note:
- This changes the "killable" behavior, the vforking task
doesn't react to the fatal signals except SIGKILL. See
the next patch
- The code asks for the final cleanups, for example we
should move put_task_struct() into clone_vfork_finish()
and simplify the usage of ->restart_block. Will be done
later
- We use ->saved_sigmask to record the original sigmask.
This is safe, nobody should play with it, we do not do.
set_restore_sigmask(). Still it would be more clean to
use restart_block->vfork, but then we should somehow
export sigset_t for thread_info.h
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
---
include/linux/thread_info.h | 4 ++++
kernel/fork.c | 37 +++++++++++++++++++++++++++++++++++--
2 files changed, 39 insertions(+), 2 deletions(-)
--- 3.1/include/linux/thread_info.h~7_vfork_restart 2011-07-27 15:27:38.000000000 +0200
+++ 3.1/include/linux/thread_info.h 2011-07-27 15:28:43.000000000 +0200
@@ -44,6 +44,10 @@ struct restart_block {
unsigned long tv_sec;
unsigned long tv_nsec;
} poll;
+
+ struct {
+ long pid;
+ } vfork;
};
};
--- 3.1/kernel/fork.c~7_vfork_restart 2011-07-27 15:27:38.000000000 +0200
+++ 3.1/kernel/fork.c 2011-07-27 16:01:01.000000000 +0200
@@ -1454,18 +1454,24 @@ static void complete_vfork_done(struct t
complete(vfork_done);
}
+static long clone_vfork_restart(struct restart_block *);
+
static long clone_vfork_finish(struct task_struct *child, long pid)
{
+ struct restart_block *restart = ¤t_thread_info()->restart_block;
struct completion vfork_done;
int killed;
+ if (!child || child->real_parent != current)
+ goto done;
+
init_completion(&vfork_done);
/* complete_vfork_done() was already called? */
if (xchg(&child->vfork_done, &vfork_done) == NULL)
goto done;
- killed = wait_for_completion_killable(&vfork_done);
+ killed = wait_for_completion_interruptible(&vfork_done);
if (killed) {
struct completion *steal = xchg(&child->vfork_done,
VFORK_DONE_NOP);
@@ -1473,14 +1479,40 @@ static long clone_vfork_finish(struct ta
if (unlikely(!steal))
wait_for_completion(&vfork_done);
- return -EINTR;
+ restart->fn = clone_vfork_restart;
+ restart->vfork.pid = pid;
+
+ return -ERESTART_RESTARTBLOCK;
}
done:
+ restart->fn = do_no_restart_syscall; /* not really needed */
+ set_current_blocked(¤t->saved_sigmask);
ptrace_event(PTRACE_EVENT_VFORK_DONE, pid);
return pid;
}
+static long clone_vfork_restart(struct restart_block *restart)
+{
+ long pid = restart->vfork.pid;
+ struct task_struct *child = find_get_task_by_vpid(pid);
+ long ret;
+
+ ret = clone_vfork_finish(child, pid);
+ if (child)
+ put_task_struct(child);
+ return ret;
+}
+
+static void clone_vfork_prepare(void)
+{
+ sigset_t vfork_mask;
+
+ current->saved_sigmask = current->blocked;
+ siginitsetinv(&vfork_mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
+ set_current_blocked(&vfork_mask);
+}
+
/*
* Ok, this is the main fork-routine.
*
@@ -1567,6 +1599,7 @@ long do_fork(unsigned long clone_flags,
ptrace_event(trace, nr);
if (clone_flags & CLONE_VFORK) {
+ clone_vfork_prepare();
nr = clone_vfork_finish(p, nr);
put_task_struct(p);
}
next prev parent reply other threads:[~2011-07-27 16:37 UTC|newest]
Thread overview: 54+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-07-27 16:31 [PATCH 0/8] make vfork killable/restartable/traceable Oleg Nesterov
2011-07-27 16:32 ` [PATCH 1/8] vfork: introduce complete_vfork_done() Oleg Nesterov
2011-07-27 16:32 ` [PATCH 2/8] vfork: introduce clone_vfork_finish() Oleg Nesterov
2011-07-27 16:32 ` [PATCH 3/8] vfork: make it killable Oleg Nesterov
2011-07-29 13:02 ` Matt Fleming
2011-07-29 14:32 ` Oleg Nesterov
2011-07-29 15:32 ` Matt Fleming
2011-07-27 16:33 ` [PATCH 4/8] coredump_wait: don't call complete_vfork_done() Oleg Nesterov
2011-07-29 13:02 ` Matt Fleming
2011-07-29 14:25 ` Oleg Nesterov
2011-07-29 15:26 ` Matt Fleming
2011-07-27 16:33 ` [PATCH 5/8] introduce find_get_task_by_vpid() Oleg Nesterov
2011-07-27 16:33 ` [PATCH 6/8] vfork: do not setup child->vfork_done beforehand Oleg Nesterov
2011-07-27 16:34 ` Oleg Nesterov [this message]
2011-07-27 16:34 ` [PATCH 8/8] vfork: do not block SIG_DFL/SIG_IGN signals is single-threaded Oleg Nesterov
2011-07-27 16:34 ` [PATCH 9/8] kill PF_STARTING Oleg Nesterov
2011-07-27 19:39 ` [PATCH 0/8] make vfork killable/restartable/traceable Linus Torvalds
2011-07-28 13:59 ` Oleg Nesterov
2011-07-28 14:58 ` Oleg Nesterov
2011-07-27 22:38 ` Pedro Alves
2011-07-29 19:23 ` Tejun Heo
2011-08-12 17:55 ` [PATCH v2 0/3] make vfork killable Oleg Nesterov
2011-08-12 17:56 ` [PATCH 1/3] vfork: introduce complete_vfork_done() Oleg Nesterov
2011-08-12 17:56 ` [PATCH 2/3] vfork: make it killable Oleg Nesterov
2011-08-19 20:33 ` Matt Fleming
2011-08-22 13:35 ` Oleg Nesterov
2011-08-12 17:56 ` [PATCH 3/3] coredump_wait: don't call complete_vfork_done() Oleg Nesterov
2011-08-17 7:50 ` Tejun Heo
2011-08-17 15:11 ` Oleg Nesterov
2011-08-12 17:57 ` [PATCH 4/3] kill PF_STARTING Oleg Nesterov
2011-08-17 7:51 ` Tejun Heo
2011-08-13 16:18 ` [PATCH v2 0/3] make vfork killable Tejun Heo
2011-08-15 19:42 ` Oleg Nesterov
2011-08-16 19:42 ` Tejun Heo
2011-08-23 22:01 ` Matt Helsley
2011-08-23 22:12 ` Tejun Heo
[not found] ` <20110727163610.GJ23793@redhat.com>
[not found] ` <20110727175624.GA3950@redhat.com>
[not found] ` <20110728154324.GA22864@redhat.com>
[not found] ` <alpine.DEB.2.00.1107281341060.16093@chino.kir.corp.google.com>
[not found] ` <20110729141431.GA3501@redhat.com>
[not found] ` <20110730143426.GA6061@redhat.com>
2011-07-30 15:22 ` mm->oom_disable_count is broken Oleg Nesterov
2011-08-01 11:52 ` KOSAKI Motohiro
2011-08-29 18:37 ` Oleg Nesterov
2011-08-29 23:17 ` David Rientjes
2011-08-30 7:43 ` [patch 1/2] oom: remove oom_disable_count David Rientjes
2011-08-30 7:43 ` David Rientjes
2011-08-30 7:43 ` [patch 2/2] oom: fix race while temporarily setting current's oom_score_adj David Rientjes
2011-08-30 7:43 ` David Rientjes
2011-08-30 15:57 ` Oleg Nesterov
2011-08-30 15:57 ` Oleg Nesterov
2011-08-30 15:28 ` [patch 1/2] oom: remove oom_disable_count Oleg Nesterov
2011-08-30 15:28 ` Oleg Nesterov
2011-08-30 22:06 ` David Rientjes
2011-08-30 22:06 ` David Rientjes
2011-08-30 16:17 ` mm->oom_disable_count is broken Oleg Nesterov
2011-08-10 21:44 ` [PATCH 0/8] make vfork killable/restartable/traceable Pavel Machek
2011-08-11 16:09 ` Oleg Nesterov
2011-08-11 16:22 ` Tejun Heo
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20110727163412.GG23793@redhat.com \
--to=oleg@redhat.com \
--cc=dvlasenk@redhat.com \
--cc=kosaki.motohiro@jp.fujitsu.com \
--cc=linux-kernel@vger.kernel.org \
--cc=matt.fleming@linux.intel.com \
--cc=roland@hack.frob.com \
--cc=tj@kernel.org \
--cc=torvalds@linux-foundation.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.