| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Fri, 2 Dec 2011 14:44:41 +0900
From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>
To: Dave Chinner <david@...morbit.com>
Cc: David Rientjes <rientjes@...gle.com>, linux-kernel@...r.kernel.org,
linux-mm@...ck.org
Subject: Re: [3.2-rc3] OOM killer doesn't kill the obvious memory hog
On Fri, 2 Dec 2011 14:31:48 +1100
Dave Chinner <david@...morbit.com> wrote:
> So, it's a distro bug - sshd should never be started from from udev
> context because of this inherited oom_score_adj thing.
> Interestingly, the ifup ssh restart script says this:
>
> # We'd like to use 'reload' here, but it has some problems; see #502444.
> if [ -x /usr/sbin/invoke-rc.d ]; then
> invoke-rc.d ssh restart >/dev/null 2>&1 || true
> else
> /etc/init.d/ssh restart >/dev/null 2>&1 || true
> fi
>
> Bug 502444 describes the exact startup race condition that I've just
> found. It does a ssh server restart because reload causes the sshd
> server to fail to start if a start is currently in progress. So,
> rather than solving the start vs reload race condition, it got a
> bandaid (use restart to restart sshd from the reload context) and
> left it as a landmine.....
>
Thank you for chasing.
Hm, BTW, do you think this kind of tracepoint is useful for debugging ?
This patch is just an example.
==
>From ed565cbf842e0b30827fba7bfdbc724fe21d9d2d Mon Sep 17 00:00:00 2001
From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>
Date: Fri, 2 Dec 2011 14:10:51 +0900
Subject: [PATCH] oom_score_adj trace point.
oom_score_adj is set by some daemon and launch tasks ans inherited
to applications, sometimes unexpectedly.
This patch is for debugging oom_score_adj inheritance. This
adds trace points for oom_score_adj inheritance.
bash-2501 [002] 448.860197: oom_score_adj_update: task 2501[bash] updates oom_score_adj=-1000
bash-2501 [002] 455.678190: oom_score_adj_inherited: new task 2527 inherited oom_score_adj -1000
ls-2527 [007] 455.678683: oom_score_task_rename: task 2527[bash] to [ls] oom_score_adj=-1000
bash-2501 [007] 461.632103: oom_score_adj_inherited: new task 2528 inherited oom_score_adj -1000
bash-2501 [007] 461.632335: oom_score_adj_inherited: new task 2529 inherited oom_score_adj -1000
ls-2528 [003] 461.632983: oom_score_task_rename: task 2528[bash] to [ls] oom_score_adj=-1000
less-2529 [005] 461.633086: oom_score_task_rename: task 2529[bash] to [less] oom_score_adj=-1000
bash-2501 [004] 474.888710: oom_score_adj_update: task 2501[bash] updates oom_score_adj=0
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>
---
fs/exec.c | 5 +++
fs/proc/base.c | 3 ++
include/trace/events/oom.h | 80 ++++++++++++++++++++++++++++++++++++++++++++
kernel/fork.c | 5 +++
mm/oom_kill.c | 6 +++
5 files changed, 99 insertions(+), 0 deletions(-)
create mode 100644 include/trace/events/oom.h
diff --git a/fs/exec.c b/fs/exec.c
index 3625464..340760f 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -59,6 +59,8 @@
#include <asm/uaccess.h>
#include <asm/mmu_context.h>
#include <asm/tlb.h>
+
+#include <trace/events/oom.h>
#include "internal.h"
int core_uses_pid;
@@ -1054,6 +1056,9 @@ void set_task_comm(struct task_struct *tsk, char *buf)
{
task_lock(tsk);
+ if (tsk->signal->oom_score_adj)
+ trace_oom_score_task_rename(tsk, buf);
+
/*
* Threads may access current->comm without holding
* the task lock, so write the string carefully.
diff --git a/fs/proc/base.c b/fs/proc/base.c
index 1050b1c..f201e64 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -87,6 +87,7 @@
#ifdef CONFIG_HARDWALL
#include <asm/hardwall.h>
#endif
+#include <trace/events/oom.h>
#include "internal.h"
/* NOTE:
@@ -1166,6 +1167,7 @@ static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
else
task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
-OOM_DISABLE;
+ trace_oom_score_adj_update(task);
err_sighand:
unlock_task_sighand(task, &flags);
err_task_lock:
@@ -1253,6 +1255,7 @@ static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
task->signal->oom_score_adj = oom_score_adj;
if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
task->signal->oom_score_adj_min = oom_score_adj;
+ trace_oom_score_adj_update(task);
/*
* Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
* always attainable.
diff --git a/include/trace/events/oom.h b/include/trace/events/oom.h
new file mode 100644
index 0000000..e161ae5
--- /dev/null
+++ b/include/trace/events/oom.h
@@ -0,0 +1,80 @@
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM oom
+
+#if !defined(_TRACE_OOM_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_OOM_H
+#include <linux/tracepoint.h>
+
+TRACE_EVENT(oom_score_adj_inherited,
+
+ TP_PROTO(struct task_struct *task),
+
+ TP_ARGS(task),
+
+ TP_STRUCT__entry(
+ __field( pid_t, newpid)
+ __field( int, oom_score_adj)
+ ),
+
+ TP_fast_assign(
+ __entry->newpid = task->pid;
+ __entry->oom_score_adj = task->signal->oom_score_adj;
+ ),
+
+ TP_printk("new task %d inherited oom_score_adj %d",
+ __entry->newpid, __entry->oom_score_adj)
+);
+
+TRACE_EVENT(oom_score_task_rename,
+
+ TP_PROTO(struct task_struct *task, char *comm),
+
+ TP_ARGS(task, comm),
+
+ TP_STRUCT__entry(
+ __field( pid_t, pid)
+ __array( char, oldcomm, TASK_COMM_LEN )
+ __array( char, newcomm, TASK_COMM_LEN )
+ __field( int, oom_score_adj)
+ ),
+
+ TP_fast_assign(
+ __entry->pid = task->pid;
+ memcpy(__entry->oldcomm, task->comm, TASK_COMM_LEN);
+ memcpy(__entry->newcomm, comm, TASK_COMM_LEN);
+ __entry->oom_score_adj = task->signal->oom_score_adj;
+ ),
+
+ TP_printk("task %d[%s] to [%s] oom_score_adj=%d",
+ __entry->pid, __entry->oldcomm, __entry->newcomm,
+ __entry->oom_score_adj)
+);
+
+TRACE_EVENT(oom_score_adj_update,
+
+ TP_PROTO(struct task_struct *task),
+
+ TP_ARGS(task),
+
+ TP_STRUCT__entry(
+ __field( pid_t, pid)
+ __array( char, comm, TASK_COMM_LEN )
+ __field( int, oom_score_adj)
+ ),
+
+ TP_fast_assign(
+ __entry->pid = task->pid;
+ memcpy(__entry->comm, task->comm, TASK_COMM_LEN);
+ __entry->oom_score_adj = task->signal->oom_score_adj;
+ ),
+
+ TP_printk("task %d[%s] updates oom_score_adj=%d",
+ __entry->pid, __entry->comm, __entry->oom_score_adj)
+);
+
+#endif
+
+/* This part must be outside protection */
+#include <trace/define_trace.h>
+
+
diff --git a/kernel/fork.c b/kernel/fork.c
index e20518d..634aa84 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -76,6 +76,7 @@
#include <asm/tlbflush.h>
#include <trace/events/sched.h>
+#include <trace/events/oom.h>
/*
* Protected counters by write_lock_irq(&tasklist_lock)
@@ -1390,6 +1391,10 @@ static struct task_struct *copy_process(unsigned long clone_flags,
if (clone_flags & CLONE_THREAD)
threadgroup_fork_read_unlock(current);
perf_event_fork(p);
+
+ if (!(clone_flags & CLONE_THREAD) && p->signal->oom_score_adj)
+ trace_oom_score_adj_inherited(p);
+
return p;
bad_fork_free_pid:
diff --git a/mm/oom_kill.c b/mm/oom_kill.c
index 3134ee2..5c8b2aa 100644
--- a/mm/oom_kill.c
+++ b/mm/oom_kill.c
@@ -33,6 +33,10 @@
#include <linux/security.h>
#include <linux/ptrace.h>
#include <linux/freezer.h>
+#include <linux/ftrace.h>
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/oom.h>
int sysctl_panic_on_oom;
int sysctl_oom_kill_allocating_task;
@@ -55,6 +59,7 @@ void compare_swap_oom_score_adj(int old_val, int new_val)
spin_lock_irq(&sighand->siglock);
if (current->signal->oom_score_adj == old_val)
current->signal->oom_score_adj = new_val;
+ trace_oom_score_adj_update(current);
spin_unlock_irq(&sighand->siglock);
}
@@ -74,6 +79,7 @@ int test_set_oom_score_adj(int new_val)
spin_lock_irq(&sighand->siglock);
old_val = current->signal->oom_score_adj;
current->signal->oom_score_adj = new_val;
+ trace_oom_score_adj_update(current);
spin_unlock_irq(&sighand->siglock);
return old_val;
--
1.7.4.1
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists