On Fri, 2013-05-24 at 10:23 -0400, Dave Jones wrote: > .config: http://paste.fedoraproject.org/14281/94052971/raw/ > > trace shows the problem process was 'cc1', so I was likely building a kernel > at the time. There was also a trinity run going on in the background. > > cmdline: nothing special.. > > BOOT_IMAGE=/vmlinuz-3.10.0-rc2+ > root=UUID=bee21cd9-1852-4d1d-9e9d-2e44332b8df1 ro rd.md=0 rd.lvm=0 rd.dm=0 > vconsole.keymap=us rd.luks=0 vconsole.font=latarcyrheb-sun16 consoleblank=0 > console=ttyUSB0,115200 console=tty0 pause_on_oops=30 audit=0 >
OK, I'm able to reproduce this. Looks like the same issue happens from other ways from userspace into the kernel (the next way was tracing system calls). Forget the other patch. Here's a new patch that is more specific to tracing and context tracking. Can you try this one out, please. -- Steve diff --git a/include/linux/preempt.h b/include/linux/preempt.h index 87a03c7..f5d4723 100644 --- a/include/linux/preempt.h +++ b/include/linux/preempt.h @@ -33,9 +33,25 @@ do { \ preempt_schedule(); \ } while (0) +#ifdef CONFIG_CONTEXT_TRACKING + +void preempt_schedule_context(void); + +#define preempt_check_resched_context() \ +do { \ + if (unlikely(test_thread_flag(TIF_NEED_RESCHED))) \ + preempt_schedule_context(); \ +} while (0) +#else + +#define preempt_check_resched_context() preempt_check_resched() + +#endif /* CONFIG_CONTEXT_TRACKING */ + #else /* !CONFIG_PREEMPT */ #define preempt_check_resched() do { } while (0) +#define preempt_check_resched_context() do { } while (0) #endif /* CONFIG_PREEMPT */ @@ -88,7 +104,7 @@ do { \ do { \ preempt_enable_no_resched_notrace(); \ barrier(); \ - preempt_check_resched(); \ + preempt_check_resched_context(); \ } while (0) #else /* !CONFIG_PREEMPT_COUNT */ diff --git a/kernel/context_tracking.c b/kernel/context_tracking.c index 65349f0..ac3a312 100644 --- a/kernel/context_tracking.c +++ b/kernel/context_tracking.c @@ -71,6 +71,46 @@ void user_enter(void) local_irq_restore(flags); } +/** + * preempt_schedule_context - preempt_schedule called by tracing + * + * The tracing infrastructure uses preempt_enable_notrace to prevent + * recursion and tracing preempt enabling caused by the tracing + * infrastructure itself. But as tracing can happen in areas coming + * from userspace or just about to enter userspace, a preempt enable + * can occur before user_exit() is called. This will cause the scheduler + * to be called when the system is still in usermode. + * + * To prevent this, the preempt_enable_notrace will use this function + * instead of preempt_schedule() to exit user context if needed before + * calling the scheduler. + */ +void __sched notrace preempt_schedule_context(void) +{ + struct thread_info *ti = current_thread_info(); + enum ctx_state prev_ctx; + + if (likely(ti->preempt_count || irqs_disabled())) + return; + + /* + * Need to disable preemption in case user_exit() is traced + * and the tracer calls preempt_enable_notrace() causing + * an infinite recursion. + */ + preempt_disable_notrace(); + prev_ctx = this_cpu_read(context_tracking.state); + user_exit(); + preempt_enable_no_resched_notrace(); + + preempt_schedule(); + + preempt_disable_notrace(); + if (prev_ctx == IN_USER) + user_enter(); + preempt_enable_notrace(); +} +EXPORT_SYMBOL(preempt_schedule_context); /** * user_exit - Inform the context tracking that the CPU is -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/