Messages in this thread Patch in this message | | | Date | Thu, 26 Mar 2009 10:40:47 -0400 (EDT) | From | Steven Rostedt <> | Subject | [PATCH][GIT PULL] tracing/wakeup: move access to wakeup_cpu into spinlock |
| |
Ingo,
I believe this is the fix for the oops that Maneesh saw. What he explains he did sounds like it would trigger the race. Reading the trace output resets wakeup_task to NULL and wakeup_cpu to -1. And he hit the bug by just reading the trace in a while loop.
Please pull the latest tip/tracing/ftrace-1 tree, which can be found at:
git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-2.6-trace.git tip/tracing/ftrace-1
Steven Rostedt (1): tracing/wakeup: move access to wakeup_cpu into spinlock
---- kernel/trace/trace_sched_wakeup.c | 6 +++--- 1 files changed, 3 insertions(+), 3 deletions(-) --------------------------- commit 8ac070ca952d14c57e3d772e203fd117161cc596 Author: Steven Rostedt <srostedt@redhat.com> Date: Thu Mar 26 10:25:24 2009 -0400
tracing/wakeup: move access to wakeup_cpu into spinlock Impact: fix for race condition The code had the following outside the lock: if (next != wakeup_task) return; pc = preempt_count(); /* The task we are waiting for is waking up */ data = wakeup_trace->data[wakeup_cpu]; On initialization, wakeup_task is NULL and wakeup_cpu -1. This code is not under a lock. If wakeup_task is set on another CPU as that task is waking up, we can see the wakeup_task before wakeup_cpu is set. If we read wakeup_cpu while it is still -1 then we will have a bad data pointer. This patch moves the reading of wakeup_cpu within the protection of the spinlock used to protect the writing of wakeup_cpu and wakeup_task. Reported-by: Maneesh Soni <maneesh@in.ibm.com> Signed-off-by: Steven Rostedt <srostedt@redhat.com>
diff --git a/kernel/trace/trace_sched_wakeup.c b/kernel/trace/trace_sched_wakeup.c index 3c5ad6b..9e4ce4c 100644 --- a/kernel/trace/trace_sched_wakeup.c +++ b/kernel/trace/trace_sched_wakeup.c @@ -138,9 +138,6 @@ probe_wakeup_sched_switch(struct rq *rq, struct task_struct *prev, pc = preempt_count(); - /* The task we are waiting for is waking up */ - data = wakeup_trace->data[wakeup_cpu]; - /* disable local data, not wakeup_cpu data */ cpu = raw_smp_processor_id(); disabled = atomic_inc_return(&wakeup_trace->data[cpu]->disabled); @@ -154,6 +151,9 @@ probe_wakeup_sched_switch(struct rq *rq, struct task_struct *prev, if (unlikely(!tracer_enabled || next != wakeup_task)) goto out_unlock; + /* The task we are waiting for is waking up */ + data = wakeup_trace->data[wakeup_cpu]; + trace_function(wakeup_trace, CALLER_ADDR1, CALLER_ADDR2, flags, pc); tracing_sched_switch_trace(wakeup_trace, prev, next, flags, pc);
| |