On Wed, Sep 08, 2021 at 12:14:23PM +0200, Peter Zijlstra wrote:
> On Tue, Sep 07, 2021 at 03:20:44PM +0200, Maarten Lankhorst wrote:
> > i915 will soon gain an eviction path that trylock a whole lot of locks
> > for eviction, getting dmesg failures like below:
> >
> > BUG: MAX_LOCK_DEPTH too low!
> > turning off the locking correctness validator.
> > depth: 48 max: 48!
> > 48 locks held by i915_selftest/5776:
> > #0: ffff888101a79240 (&dev->mutex){....}-{3:3}, at: __driver_attach+0x88/0x160
> > #1: ffffc900009778c0 (reservation_ww_class_acquire){+.+.}-{0:0}, at: i915_vma_pin.constprop.63+0x39/0x1b0 [i915]
> > #2: ffff88800cf74de8 (reservation_ww_class_mutex){+.+.}-{3:3}, at: i915_vma_pin.constprop.63+0x5f/0x1b0 [i915]
> > #3: ffff88810c7f9e38 (&vm->mutex/1){+.+.}-{3:3}, at: i915_vma_pin_ww+0x1c4/0x9d0 [i915]
> > #4: ffff88810bad5768 (reservation_ww_class_mutex){+.+.}-{3:3}, at: i915_gem_evict_something+0x110/0x860 [i915]
> > #5: ffff88810bad60e8 (reservation_ww_class_mutex){+.+.}-{3:3}, at: i915_gem_evict_something+0x110/0x860 [i915]
> > ...
> > #46: ffff88811964d768 (reservation_ww_class_mutex){+.+.}-{3:3}, at: i915_gem_evict_something+0x110/0x860 [i915]
> > #47: ffff88811964e0e8 (reservation_ww_class_mutex){+.+.}-{3:3}, at: i915_gem_evict_something+0x110/0x860 [i915]
> > INFO: lockdep is turned off.
>
> > As an intermediate solution, add an acquire context to ww_mutex_trylock,
> > which allows us to do proper nesting annotations on the trylocks, making
> > the above lockdep splat disappear.
>
> Fair enough I suppose.
What's maybe missing from the commit message
- we'll probably use this for ttm too eventually
- even when we add full ww_mutex locking we'll still have the trylock
fastpath. This is because we have a lock inversion against list locks in
these eviction paths, and the slow path unroll to drop that list lock is
a bit nasty (and defintely expensive).
iow even long term this here is needed in some form I think.
-Daniel
>
> > +/**
> > + * ww_mutex_trylock - tries to acquire the w/w mutex with optional acquire context
> > + * @lock: mutex to lock
> > + * @ctx: optional w/w acquire context
> > + *
> > + * Trylocks a mutex with the optional acquire context; no deadlock detection is
> > + * possible. Returns 1 if the mutex has been acquired successfully, 0 otherwise.
> > + *
> > + * Unlike ww_mutex_lock, no deadlock handling is performed. However, if a @ctx is
> > + * specified, -EALREADY and -EDEADLK handling may happen in calls to ww_mutex_lock.
> > + *
> > + * A mutex acquired with this function must be released with ww_mutex_unlock.
> > + */
> > +int __sched
> > +ww_mutex_trylock(struct ww_mutex *ww, struct ww_acquire_ctx *ctx)
> > +{
> > + bool locked;
> > +
> > + if (!ctx)
> > + return mutex_trylock(&ww->base);
> > +
> > +#ifdef CONFIG_DEBUG_MUTEXES
> > + DEBUG_LOCKS_WARN_ON(ww->base.magic != &ww->base);
> > +#endif
> > +
> > + preempt_disable();
> > + locked = __mutex_trylock(&ww->base);
> > +
> > + if (locked) {
> > + ww_mutex_set_context_fastpath(ww, ctx);
> > + mutex_acquire_nest(&ww->base.dep_map, 0, 1, &ctx->dep_map, _RET_IP_);
> > + }
> > + preempt_enable();
> > +
> > + return locked;
> > +}
> > +EXPORT_SYMBOL(ww_mutex_trylock);
>
> You'll need a similar hunk in ww_rt_mutex.c
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
