On Tue, Sep 23, 2014 at 05:06:41PM +0200, Peter Zijlstra wrote:
> On Mon, Sep 22, 2014 at 10:36:18PM +0400, Kirill Tkhai wrote:
> > From: Kirill Tkhai <ktkhai@xxxxxxxxxxxxx>
> >
> > Architectures, which define __ARCH_WANT_UNLOCKED_CTXSW,
> > may pull a task when it's in the middle of schedule().
> >
> > CPU1(task1 calls schedule) CPU2
> > ... schedule()
> > ... idle_balance()
> > ... load_balance()
> > ... ...
> > schedule() ...
> > prepare_lock_switch() ...
> > raw_spin_unlock(&rq1->lock) ...
> > ... raw_spin_lock(&rq1->lock)
> > ... detach_tasks();
> > ... can_migrate_task(task1)
> > ... attach_tasks(); <--- move task1 to rq2
> > ... raw_spin_unlock(&rq1->lock)
> > ... context_switch() <--- switch to task1's stack
> > ... ...
> > (using task1's stack) (using task1's stack)
> > ... ...
> > context_switch() ...
> >
> >
> > Parallel use of a single stack is not a good idea.
>
> Indeed it is, but how about we do this instead?
>
> ---
> Subject: sched,mips,ia64: Remove __ARCH_WANT_UNLOCKED_CTXSW
>
> Kirill found that there's a subtle race in the
> __ARCH_WANT_UNLOCKED_CTXSW code, and instead of fixing it, remove the
> entire exception because neither arch that uses it seems to actually
> still require it.
>
> Boot tested on mips64el (qemu) only.
>
> Cc: Oleg Nesterov <oleg@xxxxxxxxxx>
> Cc: Guenter Roeck <linux@xxxxxxxxxxxx>
> Cc: Ralf Baechle <ralf@xxxxxxxxxxxxxx>
> Cc: Tony Luck <tony.luck@xxxxxxxxx>
> Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
> ---
> arch/ia64/include/asm/processor.h | 1 -
> arch/mips/include/asm/processor.h | 6 ------
> kernel/sched/core.c | 6 ------
> kernel/sched/sched.h | 30 ------------------------------
You might also want to clean up any references to the define if you remove it.
Documentation/scheduler/sched-arch.txt:you must `#define __ARCH_WANT_UNLOCKED_CTXSW` in a header file
arch/blackfin/kernel/entry.S: * since Blackfin does not define __ARCH_WANT_UNLOCKED_CTXSW, so
Guenter
> 4 files changed, 43 deletions(-)
>
> diff --git a/arch/ia64/include/asm/processor.h b/arch/ia64/include/asm/processor.h
> index c736713..ce53c50 100644
> --- a/arch/ia64/include/asm/processor.h
> +++ b/arch/ia64/include/asm/processor.h
> @@ -19,7 +19,6 @@
> #include <asm/ptrace.h>
> #include <asm/ustack.h>
>
> -#define __ARCH_WANT_UNLOCKED_CTXSW
> #define ARCH_HAS_PREFETCH_SWITCH_STACK
>
> #define IA64_NUM_PHYS_STACK_REG 96
> diff --git a/arch/mips/include/asm/processor.h b/arch/mips/include/asm/processor.h
> index 05f0843..f1df4cb 100644
> --- a/arch/mips/include/asm/processor.h
> +++ b/arch/mips/include/asm/processor.h
> @@ -397,12 +397,6 @@ unsigned long get_wchan(struct task_struct *p);
> #define ARCH_HAS_PREFETCHW
> #define prefetchw(x) __builtin_prefetch((x), 1, 1)
>
> -/*
> - * See Documentation/scheduler/sched-arch.txt; prevents deadlock on SMP
> - * systems.
> - */
> -#define __ARCH_WANT_UNLOCKED_CTXSW
> -
> #endif
>
> #endif /* _ASM_PROCESSOR_H */
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 2a93b87..ccbafb0 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -2304,10 +2304,6 @@ asmlinkage __visible void schedule_tail(struct task_struct *prev)
> */
> post_schedule(rq);
>
> -#ifdef __ARCH_WANT_UNLOCKED_CTXSW
> - /* In this case, finish_task_switch does not reenable preemption */
> - preempt_enable();
> -#endif
> if (current->set_child_tid)
> put_user(task_pid_vnr(current), current->set_child_tid);
> }
> @@ -2350,9 +2346,7 @@ context_switch(struct rq *rq, struct task_struct *prev,
> * of the scheduler it's an obvious special-case), so we
> * do an early lockdep release here:
> */
> -#ifndef __ARCH_WANT_UNLOCKED_CTXSW
> spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
> -#endif
>
> context_tracking_task_switch(prev, next);
> /* Here we just switch the register state and the stack. */
> diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
> index 1bc6aad..d87f122 100644
> --- a/kernel/sched/sched.h
> +++ b/kernel/sched/sched.h
> @@ -966,7 +966,6 @@ static inline int task_on_rq_migrating(struct task_struct *p)
> # define finish_arch_post_lock_switch() do { } while (0)
> #endif
>
> -#ifndef __ARCH_WANT_UNLOCKED_CTXSW
> static inline void prepare_lock_switch(struct rq *rq, struct task_struct *next)
> {
> #ifdef CONFIG_SMP
> @@ -1004,35 +1003,6 @@ static inline void finish_lock_switch(struct rq *rq, struct task_struct *prev)
> raw_spin_unlock_irq(&rq->lock);
> }
>
> -#else /* __ARCH_WANT_UNLOCKED_CTXSW */
> -static inline void prepare_lock_switch(struct rq *rq, struct task_struct *next)
> -{
> -#ifdef CONFIG_SMP
> - /*
> - * We can optimise this out completely for !SMP, because the
> - * SMP rebalancing from interrupt is the only thing that cares
> - * here.
> - */
> - next->on_cpu = 1;
> -#endif
> - raw_spin_unlock(&rq->lock);
> -}
> -
> -static inline void finish_lock_switch(struct rq *rq, struct task_struct *prev)
> -{
> -#ifdef CONFIG_SMP
> - /*
> - * After ->on_cpu is cleared, the task can be moved to a different CPU.
> - * We must ensure this doesn't happen until the switch is completely
> - * finished.
> - */
> - smp_wmb();
> - prev->on_cpu = 0;
> -#endif
> - local_irq_enable();
> -}
> -#endif /* __ARCH_WANT_UNLOCKED_CTXSW */
> -
> /*
> * wake flags
> */
