From: Oleg Nesterov <oleg@redhat.com>
To: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: akpm@linux-foundation.org, peterz@infradead.org,
viro@zeniv.linux.org.uk, mingo@kernel.org,
paulmck@linux.vnet.ibm.com, keescook@chromium.org,
riel@redhat.com, mhocko@suse.com, tglx@linutronix.de,
kirill.shutemov@linux.intel.com, marcos.souza.org@gmail.com,
hoeun.ryu@gmail.com, pasha.tatashin@oracle.com,
gs051095@gmail.com, ebiederm@xmission.com, dhowells@redhat.com,
rppt@linux.vnet.ibm.com, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 1/4] exit: Move read_unlock() up in mm_update_next_owner()
Date: Thu, 26 Apr 2018 17:01:17 +0200 [thread overview]
Message-ID: <20180426150116.GA14818@redhat.com> (raw)
In-Reply-To: <152474043375.29458.13978538538182642678.stgit@localhost.localdomain>
On 04/26, Kirill Tkhai wrote:
>
> @@ -464,18 +464,15 @@ void mm_update_next_owner(struct mm_struct *mm)
> return;
>
> assign_new_owner:
> - BUG_ON(c == p);
> get_task_struct(c);
> + read_unlock(&tasklist_lock);
> + BUG_ON(c == p);
> +
> /*
> * The task_lock protects c->mm from changing.
> * We always want mm->owner->mm == mm
> */
> task_lock(c);
> - /*
> - * Delay read_unlock() till we have the task_lock()
> - * to ensure that c does not slip away underneath us
> - */
> - read_unlock(&tasklist_lock);
I think this is correct, but...
Firstly, I agree with Michal, it would be nice to kill mm_update_next_owner()
altogether.
If this is not possible I agree, it needs cleanups and we can change it to
avoid tasklist (although your 4/4 looks overcomplicated to me at first glance).
But in this case I think that whatever we do we should start with something like
the patch below. I wrote it 3 years ago but it still applies.
Oleg.
Subject: [PATCH 1/3] memcg: introduce assign_new_owner()
The code under "assign_new_owner" looks very ugly and suboptimal.
We do not really need get_task_struct/put_task_struct(), we can
simply recheck/change mm->owner under tasklist_lock. And we do not
want to restart from the very beginning if ->mm was changed by the
time we take task_lock(), we can simply continue (if we do not drop
tasklist_lock).
Just move this code into the new simple helper, assign_new_owner().
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
---
kernel/exit.c | 56 ++++++++++++++++++++++++++------------------------------
1 files changed, 26 insertions(+), 30 deletions(-)
diff --git a/kernel/exit.c b/kernel/exit.c
index 22fcc05..4d446ab 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -293,6 +293,23 @@ kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
}
#ifdef CONFIG_MEMCG
+static bool assign_new_owner(struct mm_struct *mm, struct task_struct *c)
+{
+ bool ret = false;
+
+ if (c->mm != mm)
+ return ret;
+
+ task_lock(c); /* protects c->mm from changing */
+ if (c->mm == mm) {
+ mm->owner = c;
+ ret = true;
+ }
+ task_unlock(c);
+
+ return ret;
+}
+
/*
* A task is exiting. If it owned this mm, find a new owner for the mm.
*/
@@ -300,7 +317,6 @@ void mm_update_next_owner(struct mm_struct *mm)
{
struct task_struct *c, *g, *p = current;
-retry:
/*
* If the exiting or execing task is not the owner, it's
* someone else's problem.
@@ -322,16 +338,16 @@ retry:
* Search in the children
*/
list_for_each_entry(c, &p->children, sibling) {
- if (c->mm == mm)
- goto assign_new_owner;
+ if (assign_new_owner(mm, c))
+ goto done;
}
/*
* Search in the siblings
*/
list_for_each_entry(c, &p->real_parent->children, sibling) {
- if (c->mm == mm)
- goto assign_new_owner;
+ if (assign_new_owner(mm, c))
+ goto done;
}
/*
@@ -341,42 +357,22 @@ retry:
if (g->flags & PF_KTHREAD)
continue;
for_each_thread(g, c) {
- if (c->mm == mm)
- goto assign_new_owner;
+ if (assign_new_owner(mm, c))
+ goto done;
if (c->mm)
break;
}
}
- read_unlock(&tasklist_lock);
+
/*
* We found no owner yet mm_users > 1: this implies that we are
* most likely racing with swapoff (try_to_unuse()) or /proc or
* ptrace or page migration (get_task_mm()). Mark owner as NULL.
*/
mm->owner = NULL;
- return;
-
-assign_new_owner:
- BUG_ON(c == p);
- get_task_struct(c);
- /*
- * The task_lock protects c->mm from changing.
- * We always want mm->owner->mm == mm
- */
- task_lock(c);
- /*
- * Delay read_unlock() till we have the task_lock()
- * to ensure that c does not slip away underneath us
- */
+done:
read_unlock(&tasklist_lock);
- if (c->mm != mm) {
- task_unlock(c);
- put_task_struct(c);
- goto retry;
- }
- mm->owner = c;
- task_unlock(c);
- put_task_struct(c);
+ return;
}
#endif /* CONFIG_MEMCG */
--
1.5.5.1
next prev parent reply other threads:[~2018-04-26 15:01 UTC|newest]
Thread overview: 96+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-04-26 11:00 [PATCH 0/4] exit: Make unlikely case in mm_update_next_owner() more scalable Kirill Tkhai
2018-04-26 11:00 ` [PATCH 1/4] exit: Move read_unlock() up in mm_update_next_owner() Kirill Tkhai
2018-04-26 15:01 ` Oleg Nesterov [this message]
2018-04-26 11:00 ` [PATCH 2/4] exit: Use rcu instead of get_task_struct() " Kirill Tkhai
2018-04-26 11:00 ` [PATCH 3/4] exit: Rename assign_new_owner label " Kirill Tkhai
2018-04-26 11:01 ` [PATCH 4/4] exit: Lockless iteration over task list " Kirill Tkhai
2018-04-26 12:35 ` Andrea Parri
2018-04-26 13:52 ` Kirill Tkhai
2018-04-26 15:20 ` Peter Zijlstra
2018-04-26 15:56 ` Kirill Tkhai
2018-04-26 15:20 ` Peter Zijlstra
2018-04-26 16:04 ` Kirill Tkhai
2018-04-26 15:29 ` Andrea Parri
2018-04-26 16:11 ` Kirill Tkhai
2018-04-26 13:07 ` [PATCH 0/4] exit: Make unlikely case in mm_update_next_owner() more scalable Michal Hocko
2018-04-26 13:52 ` Oleg Nesterov
2018-04-26 14:07 ` Kirill Tkhai
2018-04-26 15:10 ` Oleg Nesterov
2018-04-26 16:19 ` Eric W. Biederman
2018-04-26 19:28 ` Michal Hocko
2018-04-27 7:08 ` Michal Hocko
2018-04-27 18:05 ` Eric W. Biederman
2018-05-01 17:22 ` Eric W. Biederman
2018-05-01 17:35 ` [RFC][PATCH] memcg: Replace mm->owner with mm->memcg Eric W. Biederman
2018-05-02 8:47 ` Michal Hocko
2018-05-02 13:20 ` Johannes Weiner
2018-05-02 14:05 ` Eric W. Biederman
2018-05-02 19:21 ` [PATCH] " Eric W. Biederman
2018-05-02 21:04 ` Andrew Morton
2018-05-02 21:35 ` Eric W. Biederman
2018-05-03 13:33 ` Oleg Nesterov
2018-05-03 14:39 ` Eric W. Biederman
2018-05-04 14:20 ` Oleg Nesterov
2018-05-04 14:36 ` Eric W. Biederman
2018-05-04 14:54 ` Oleg Nesterov
2018-05-04 15:49 ` Eric W. Biederman
2018-05-04 16:22 ` Oleg Nesterov
2018-05-04 16:40 ` Eric W. Biederman
2018-05-04 17:26 ` [PATCH 0/2] mm->owner to mm->memcg fixes Eric W. Biederman
2018-05-04 17:26 ` [PATCH 1/2] memcg: Update the mm->memcg maintenance to work when !CONFIG_MMU Eric W. Biederman
2018-05-04 17:27 ` [PATCH 2/2] memcg: Close the race between migration and installing bprm->mm as mm Eric W. Biederman
2018-05-09 14:51 ` Oleg Nesterov
2018-05-10 3:00 ` Eric W. Biederman
2018-05-10 12:14 ` [PATCH 0/2] mm->owner to mm->memcg fixes Michal Hocko
2018-05-10 12:18 ` Michal Hocko
2018-05-22 12:57 ` Michal Hocko
2018-05-23 19:46 ` Eric W. Biederman
2018-05-24 11:10 ` Michal Hocko
2018-05-24 21:16 ` Andrew Morton
2018-05-24 23:37 ` Andrea Parri
2018-05-30 12:17 ` Michal Hocko
2018-05-31 18:41 ` Eric W. Biederman
2018-06-01 1:57 ` [PATCH] memcg: Replace mm->owner with mm->memcg Eric W. Biederman
2018-06-01 14:52 ` [RFC][PATCH 0/2] memcg: Require every task that uses an mm to migrate together Eric W. Biederman
2018-06-01 14:53 ` [RFC][PATCH 1/2] memcg: Ensure every task that uses an mm is in the same memory cgroup Eric W. Biederman
2018-06-01 16:50 ` Tejun Heo
2018-06-01 18:11 ` Eric W. Biederman
2018-06-01 19:16 ` Tejun Heo
2018-06-04 13:01 ` Michal Hocko
2018-06-04 18:47 ` Tejun Heo
2018-06-04 19:11 ` Eric W. Biederman
2018-06-06 11:13 ` Michal Hocko
2018-06-07 11:42 ` Eric W. Biederman
2018-06-07 12:19 ` Michal Hocko
2018-06-01 14:53 ` [RFC][PATCH 2/2] memcgl: Remove dead code now that all tasks of an mm share a memcg Eric W. Biederman
2018-06-01 14:07 ` [PATCH 0/2] mm->owner to mm->memcg fixes Michal Hocko
2018-05-24 21:17 ` Andrew Morton
2018-05-30 11:52 ` Michal Hocko
2018-05-31 17:43 ` Eric W. Biederman
2018-05-07 14:33 ` [PATCH] memcg: Replace mm->owner with mm->memcg Oleg Nesterov
2018-05-08 3:15 ` Eric W. Biederman
2018-05-09 14:40 ` Oleg Nesterov
2018-05-10 3:09 ` Eric W. Biederman
2018-05-10 4:03 ` [RFC][PATCH] cgroup: Don't mess with tasks in exec Eric W. Biederman
2018-05-10 12:15 ` Oleg Nesterov
2018-05-10 12:35 ` Tejun Heo
2018-05-10 12:38 ` [PATCH] memcg: Replace mm->owner with mm->memcg Oleg Nesterov
2018-05-04 11:07 ` Michal Hocko
2018-05-05 16:54 ` kbuild test robot
2018-05-07 23:18 ` Andrew Morton
2018-05-08 2:17 ` Eric W. Biederman
2018-05-09 21:00 ` Michal Hocko
2018-05-02 23:59 ` [RFC][PATCH] " Balbir Singh
2018-05-03 15:11 ` Eric W. Biederman
2018-05-04 4:59 ` Balbir Singh
2018-05-03 10:52 ` [PATCH 0/4] exit: Make unlikely case in mm_update_next_owner() more scalable Kirill Tkhai
2018-06-01 1:07 ` Eric W. Biederman
2018-06-01 13:57 ` Michal Hocko
2018-06-01 14:32 ` Eric W. Biederman
2018-06-01 15:02 ` Michal Hocko
2018-06-01 15:25 ` Eric W. Biederman
2018-06-04 6:54 ` Michal Hocko
2018-06-04 14:31 ` Eric W. Biederman
2018-06-05 8:15 ` Michal Hocko
2018-06-05 8:48 ` Kirill Tkhai
2018-06-05 15:36 ` Eric W. Biederman
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=20180426150116.GA14818@redhat.com \
--to=oleg@redhat.com \
--cc=akpm@linux-foundation.org \
--cc=dhowells@redhat.com \
--cc=ebiederm@xmission.com \
--cc=gs051095@gmail.com \
--cc=hoeun.ryu@gmail.com \
--cc=keescook@chromium.org \
--cc=kirill.shutemov@linux.intel.com \
--cc=ktkhai@virtuozzo.com \
--cc=linux-kernel@vger.kernel.org \
--cc=marcos.souza.org@gmail.com \
--cc=mhocko@suse.com \
--cc=mingo@kernel.org \
--cc=pasha.tatashin@oracle.com \
--cc=paulmck@linux.vnet.ibm.com \
--cc=peterz@infradead.org \
--cc=riel@redhat.com \
--cc=rppt@linux.vnet.ibm.com \
--cc=tglx@linutronix.de \
--cc=viro@zeniv.linux.org.uk \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).