Op 10-03-15 om 13:37 schreef Peter Zijlstra: > On Fri, Feb 27, 2015 at 05:57:08PM +0100, Sebastian Andrzej Siewior wrote: >> +static int __sched __mutex_lock_check_stamp(struct rt_mutex *lock, >> + struct ww_acquire_ctx *ctx) >> +{ >> +#ifdef CONFIG_WW_MUTEX_RTMUTEX >> + struct ww_mutex *ww = container_of(lock, struct ww_mutex, base.lock); >> + struct ww_acquire_ctx *hold_ctx = ACCESS_ONCE(ww->ctx); >> + >> + if (!hold_ctx) >> + return 0; >> + >> + if (unlikely(ctx == hold_ctx)) >> + return -EALREADY; >> + >> + if (ctx->stamp - hold_ctx->stamp <= LONG_MAX && >> + (ctx->stamp != hold_ctx->stamp || ctx > hold_ctx)) { >> +#ifdef CONFIG_DEBUG_MUTEXES >> + DEBUG_LOCKS_WARN_ON(ctx->contending_lock); >> + ctx->contending_lock = ww; >> +#endif >> + return -EDEADLK; >> + } >> +#endif >> + return 0; >> +} > So IIRC this is the function that checks who gets wounded (and gets to > do the whole retry thing), right? > > So for the RT case, I think we should extend it to not (primarily) be a > FIFO thing, but also consider the priority of the tasks involved. > > Maybe a little something like: > > if (hold_ctx->task->prio < ctx->task->prio) > return -EDEADLOCK; > > before the timestamp check; although I suppose we should also add a > deadline test in case both prios are -1. I think that's useful but if you implement -EDEADLK based on thread priority, any boosted thread should receive -EDEADLK when it tries to acquire a new lock in the same context, to force it to back off.. I'm not sure how you'd implement it though..
~Maarten -- 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/